Linslab commited on
Commit
4cfe1b1
·
verified ·
1 Parent(s): 9d4a9e9

Add files using upload-large-folder tool

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/controller.py +230 -0
  2. agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/demo_analysis.md +83 -0
  3. agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/demo_manifest.json +1488 -0
  4. agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/findings.md +94 -0
  5. agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/geometry_clean_provenance.json +4 -0
  6. agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/geometry_provenance.json +182 -0
  7. agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/handoff.md +10 -0
  8. agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/policy_runner.py +25 -0
  9. agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/posts_inferred_provenance.json +3924 -0
  10. agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/sequences_provenance.json +122 -0
  11. agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/worker_api.py +81 -0
  12. agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/__pycache__/build_bank.cpython-310.pyc +0 -0
  13. agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/__pycache__/check_synthetic.cpython-310.pyc +0 -0
  14. agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/__pycache__/controller.cpython-310.pyc +0 -0
  15. agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/__pycache__/features.cpython-310.pyc +0 -0
  16. agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/controller.py +66 -0
  17. agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/environment.lock +249 -0
  18. agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/features.py +23 -0
  19. agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/freeze_manifest.json +142 -0
  20. agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/model_tool_manifest.yaml +69 -0
  21. agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/skills.md +4 -0
  22. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/__pycache__/build_memory.cpython-310.pyc +0 -0
  23. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/__pycache__/check_synthetic.cpython-310.pyc +0 -0
  24. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/__pycache__/controller.cpython-310.pyc +0 -0
  25. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/__pycache__/features.cpython-310.pyc +0 -0
  26. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/build_memory.py +25 -0
  27. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/check_synthetic.py +23 -0
  28. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/checks.txt +6 -0
  29. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/config.json +130 -0
  30. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/controller.py +68 -0
  31. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/demo_manifest.json +2168 -0
  32. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/environment.lock +249 -0
  33. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/features.py +28 -0
  34. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/findings.md +17 -0
  35. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/freeze_manifest.json +96 -0
  36. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/handoff.md +7 -0
  37. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/model_tool_manifest.yaml +119 -0
  38. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/policy_runner.py +25 -0
  39. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/skills.md +8 -0
  40. agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/worker_api.py +81 -0
  41. agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/controller.py +44 -0
  42. agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/demo_manifest.json +2168 -0
  43. agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/environment.lock +249 -0
  44. agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/freeze_manifest.json +117 -0
  45. agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/handoff.md +11 -0
  46. agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/kinematics.py +8 -0
  47. agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/model_tool_manifest.yaml +93 -0
  48. agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/skills.md +11 -0
  49. agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/vision.py +12 -0
  50. agent_d20_e20/frozen_20261102/tasks/DualArmMahjongSetup-v1/frozen/bank.npz +3 -0
agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/controller.py ADDED
@@ -0,0 +1,230 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import cv2,numpy as np
2
+
3
+ def posts_hough(im):
4
+ im=cv2.resize(np.asarray(im).squeeze(),(480,640),interpolation=cv2.INTER_LINEAR)
5
+ hsv=cv2.cvtColor(im,cv2.COLOR_RGB2HSV);g=cv2.cvtColor(im,cv2.COLOR_RGB2GRAY)
6
+ cs=cv2.HoughCircles(g,cv2.HOUGH_GRADIENT,1,9,param1=70,param2=8,minRadius=3,maxRadius=7)
7
+ out=[]
8
+ if cs is None:return out
9
+ for x,y,r in cs[0]:
10
+ if not 170<x<430 or not 170<y<470:continue
11
+ xx,yy=int(x),int(y)
12
+ if np.median(hsv[yy-1:yy+2,xx-1:xx+2,1])>80 or np.median(hsv[yy-1:yy+2,xx-1:xx+2,2])<165:continue
13
+ ang=np.arange(12)*2*np.pi/12;px=np.clip((x+np.cos(ang)*(r+3)).astype(int),0,479);py=np.clip((y+np.sin(ang)*(r+3)).astype(int),0,639)
14
+ if np.sum((hsv[py,px,1]>75)&(hsv[py,px,2]<220))<3:continue
15
+ out.append([float(x/2),float(y/2),float(np.pi*r*r/4)])
16
+ return out
17
+
18
+ def update_posts(state, grid, im):
19
+ groups=[[] for _ in range(4)]
20
+ wrist=wrist_hough(im)
21
+ for direct, points in [(True,posts(im)),(False,posts_hough(im))]:
22
+ for px,py,area in points:
23
+ if not 4<area<(25 if direct else 40):continue
24
+ xy=np.array([px,py])
25
+ if np.isfinite(wrist).all() and np.linalg.norm(xy-wrist)<18:continue
26
+ dist=np.linalg.norm(grid-xy,axis=1);j=int(np.argmin(dist))
27
+ if dist[j]>32 or np.sort(dist)[1]-dist[j]<5:continue
28
+ groups[j].append((xy,direct))
29
+ for j,group in enumerate(groups):
30
+ if not group:continue
31
+ anchor=state.post_positions[j] if state.post_counts[j] else grid[j]
32
+ preferred=[v for v in group if v[1]] or group
33
+ xy,direct=min(preferred,key=lambda v:np.linalg.norm(v[0]-anchor))
34
+ if state.post_counts[j]==0 or (state.post_counts[j]<5 and np.linalg.norm(xy-anchor)>2):
35
+ state.post_positions[j]=xy;state.post_counts[j]=1;state.direct_counts[j]=int(direct)
36
+ elif np.linalg.norm(xy-anchor)<=2:
37
+ state.post_positions[j]=.9*anchor+.1*xy;state.post_counts[j]+=1;state.direct_counts[j]+=int(direct)
38
+ known=np.flatnonzero(state.post_counts>=5)
39
+ best=None;best_score=-1
40
+ z=grid[:,0]+1j*grid[:,1]
41
+ for ii,j in enumerate(known):
42
+ for k in known[ii+1:]:
43
+ p=state.post_positions
44
+ c=complex(*(p[k]-p[j]))/(z[k]-z[j])
45
+ if not .92<abs(c)<1.08 or abs(np.angle(c))>.75:continue
46
+ zz=c*(z-z[j])+complex(*p[j]);pred=np.stack([zz.real,zz.imag],axis=1)
47
+ residual=np.linalg.norm(pred[known]-p[known],axis=1)
48
+ score=float(np.sum(state.post_counts[known][residual<3]))
49
+ if score>best_score:best=pred;best_score=score
50
+ if best is not None:
51
+ old_score=-1 if state.live_grid is None else np.sum(state.post_counts[known][np.linalg.norm(state.live_grid[known]-state.post_positions[known],axis=1)<3])
52
+ if best_score>=old_score:state.live_grid=best
53
+
54
+ def target_post_delta(state, grid, target):
55
+ if state.live_grid is not None:
56
+ delta=state.live_grid[target]-grid[target]
57
+ if np.linalg.norm(delta)<35:return delta,True
58
+ return np.zeros(2),False
59
+
60
+ import cv2,numpy as np
61
+
62
+ def posts(im):
63
+ im=cv2.resize(np.asarray(im).squeeze(),(240,320),interpolation=cv2.INTER_AREA)
64
+ hsv=cv2.cvtColor(im,cv2.COLOR_RGB2HSV)
65
+ mask=((hsv[:,:,1]<70)&(hsv[:,:,2]>170)).astype(np.uint8)*255
66
+ cs,_=cv2.findContours(mask,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)
67
+ out=[]
68
+ for c in cs:
69
+ a=cv2.contourArea(c);p=cv2.arcLength(c,True)
70
+ if 2<a<55 and p>0 and 4*np.pi*a/p**2>.45:
71
+ m=cv2.moments(c);x,y=m['m10']/m['m00'],m['m01']/m['m00']
72
+ if 85<x<215 and 85<y<235:out.append([x,y,a])
73
+ return out
74
+
75
+ import numpy as np,cv2
76
+
77
+ def holes(im):
78
+ im=cv2.resize(np.asarray(im).squeeze(),(240,320),interpolation=cv2.INTER_AREA)
79
+ hsv=cv2.cvtColor(im,cv2.COLOR_RGB2HSV)
80
+ white=((hsv[:,:,1]<75)&(hsv[:,:,2]>150)).astype(np.uint8)*255
81
+ cs,hh=cv2.findContours(white,cv2.RETR_CCOMP,cv2.CHAIN_APPROX_SIMPLE)
82
+ out=[]
83
+ if hh is None:return out
84
+ for c,h in zip(cs,hh[0]):
85
+ area=cv2.contourArea(c);per=cv2.arcLength(c,True);x,y,w,height=cv2.boundingRect(c)
86
+ if h[3]>=0 and 20<area<180 and per>0 and 4*np.pi*area/per**2>.65 and x>70:
87
+ m=cv2.moments(c);out.append([m['m10']/m['m00'],m['m01']/m['m00']])
88
+ return out
89
+
90
+ def wrist(im):
91
+ im=cv2.resize(np.asarray(im).squeeze(),(240,320),interpolation=cv2.INTER_AREA)
92
+ hsv=cv2.cvtColor(im,cv2.COLOR_RGB2HSV)
93
+ mask=((hsv[:,:,1]<28)&(hsv[:,:,2]>45)&(hsv[:,:,2]<180)).astype(np.uint8)*255
94
+ cs,_=cv2.findContours(mask,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)
95
+ out=[]
96
+ for c in cs:
97
+ area=cv2.contourArea(c);per=cv2.arcLength(c,True)
98
+ if 80<area<900 and per>0 and 4*np.pi*area/per**2>.6:
99
+ (x,y),r=cv2.minEnclosingCircle(c);out.append((x,y,r,4*np.pi*area/per**2))
100
+ return out
101
+
102
+ def wrist_hough(im):
103
+ im=cv2.resize(np.asarray(im).squeeze(),(240,320),interpolation=cv2.INTER_AREA)
104
+ g=cv2.cvtColor(im,cv2.COLOR_RGB2GRAY);hsv=cv2.cvtColor(im,cv2.COLOR_RGB2HSV)
105
+ cs=cv2.HoughCircles(g,cv2.HOUGH_GRADIENT,1,12,param1=80,param2=14,minRadius=6,maxRadius=13)
106
+ if cs is None:return np.array([np.nan,np.nan])
107
+ valid=[]
108
+ for x,y,r in cs[0]:
109
+ xx,yy=int(x),int(y)
110
+ if x>75 and 50<y<270 and np.median(hsv[max(0,yy-3):yy+4,max(0,xx-3):xx+4,1])<60 and 35<np.median(g[max(0,yy-3):yy+4,max(0,xx-3):xx+4])<200:valid.append((x,y))
111
+ return np.array(max(valid,key=lambda p:p[0]) if valid else [np.nan,np.nan])
112
+
113
+ """Authorized-demo trajectory alignment with current-pose correction; no learned weights."""
114
+ from pathlib import Path
115
+ import numpy as np
116
+ import cv2
117
+
118
+
119
+ def descriptor(im):
120
+ im=np.asarray(im).squeeze()
121
+ im=cv2.resize(im,(120,160),interpolation=cv2.INTER_AREA)
122
+ hsv=cv2.cvtColor(im,cv2.COLOR_RGB2HSV)
123
+ mask=((hsv[:,:,1]<55)&(hsv[:,:,2]>135)).astype(np.float32)
124
+ # Exclude the initial robot footprint from cloth localization.
125
+ mask[:,:50]=0;mask[:45]=0;mask[120:]=0
126
+ return cv2.resize(mask,(60,80),interpolation=cv2.INTER_AREA)
127
+
128
+ class Controller:
129
+ def __init__(self):
130
+ self.bank=None;self.reset_episode()
131
+ def reset_episode(self):
132
+ self.selected=None;self.t=1;self.calls=0;self.direct_counts=np.zeros(4,dtype=int);self.live_grid=None;self.last_target_delta=np.zeros(2);self.target_known=False;self.release_wait={};self.last_release_t=None;self.post_bank=None;self.post_positions=np.full((4,2),np.nan);self.post_counts=np.zeros(4,dtype=int);self.geometry=None;self.offset=np.zeros(2);self.approach_key=None;self.last_diagnostic={};self.previous=None
133
+ def act(self,obs):
134
+ self.calls+=1
135
+ step=int(obs.get('step_id',self.calls))
136
+ if step>=1000:return None
137
+ q=np.asarray(obs['proprioception']['agent/qpos']).reshape(-1)
138
+ if self.bank is None:
139
+ with np.load(Path(__file__).with_name('sequences.npz')) as z:self.bank={k:z[k] for k in z.files}
140
+ if step<25:
141
+ self.t=step+1
142
+ return np.array([0,0,0,0,0,0,1.],dtype=np.float32)
143
+ if self.post_bank is None:
144
+ with np.load(Path(__file__).with_name('posts_inferred.npz')) as z:self.post_bank={k:z[k] for k in z.files}
145
+ if self.geometry is None:
146
+ with np.load(Path(__file__).with_name('geometry_clean.npz')) as z:self.geometry={k:z[k] for k in z.files}
147
+ if self.selected is None:
148
+ f=descriptor(obs['rgb']['top_camera'])
149
+ scores=[np.mean((descriptor(self.bank[f'demo_{i:02d}_im'])-f)**2) for i in range(20)]
150
+ self.selected=f'demo_{np.argmin(scores):02d}';self.t=25
151
+ self.initial_score=float(min(scores))
152
+ a=self.bank[self.selected+'_a'];qs=self.bank[self.selected+'_q']
153
+ if self.t>=len(a):return None
154
+ # Re-align only locally by observed joint pose, with a one-frame timing prior.
155
+ lo=max(25,self.t-1);hi=min(len(a),self.t+4)
156
+ ix=np.arange(lo,hi)
157
+ costs=np.mean((qs[ix,:7]-q[:7])**2,axis=1)+.0004*(ix-self.t)**2
158
+ t=min(self.t,int(ix[np.argmin(costs)]))
159
+ transitions=np.flatnonzero(np.diff(a[:,6])!=0)+1
160
+ if np.any(np.abs(transitions-self.t)<=4):t=self.t
161
+ command=a[t].copy()
162
+ # A local finite-difference inverse maps joint tracking error to EE deltas.
163
+ lo=max(0,t-25);hi=min(len(a),t+25)
164
+ aa=a[lo:hi,:6].astype(float);dq=np.diff(qs[lo:hi+1,:7],axis=0)
165
+ J=np.linalg.solve(aa.T@aa+.002*np.eye(6),aa.T@dq)
166
+ err=qs[t,:7]-q[:7]
167
+ correction=np.linalg.solve(J@J.T+.02*np.eye(6),J@err)
168
+ command[:3]+=np.clip(correction[:3],-.005,.005)
169
+ # Keep demonstrated orientation deltas; position errors must not induce wrist rotation.
170
+ # Match the next grasp's reference eyelet to the current legal image.
171
+ ws=self.geometry[self.selected+'_w'];hs=self.geometry[self.selected+'_h']
172
+ visual_used=False
173
+ grid=self.post_bank[self.selected]
174
+ update_posts(self,grid,obs['rgb']['top_camera'])
175
+ target_delta=np.zeros(2)
176
+ future=np.flatnonzero((a[:,6]<-.5)&(np.r_[1,a[:-1,6]]>=-.5))
177
+ closes=future[future>=t]
178
+ if len(closes) and command[6]==0:
179
+ close=int(closes[0]);ww=ws[max(25,close-5):close+1];ww=ww[np.isfinite(ww).all(axis=1)]
180
+ ref=hs[t];ref=ref[np.isfinite(ref).all(axis=1)]
181
+ live=np.array(holes(obs['rgb']['top_camera']))
182
+ if len(ww) and len(ref) and len(live):
183
+ goal=ww[-1];target=ref[np.argmin(np.linalg.norm(ref-goal,axis=1))]
184
+ match=live[np.argmin(np.linalg.norm(live-target,axis=1))]
185
+ delta=match-target
186
+ if np.linalg.norm(delta)<25 and np.linalg.norm(target-goal)<40:
187
+ self.offset=.8*self.offset+.2*delta if self.approach_key==close else delta
188
+ self.approach_key=close
189
+ else:
190
+ past=future[future<=t]
191
+ dwell=(command[6]<-.5 and len(past)>0 and t-int(past[-1])<12)
192
+ if command[6]<-.5 and len(past):
193
+ self.target_known=False
194
+ close=int(past[-1]);release=np.flatnonzero(a[close:,6]>=0)
195
+ end=close+int(release[0]) if len(release) else len(a)-1
196
+ ww=ws[max(0,end-3):min(len(ws),end+2)];ww=ww[np.isfinite(ww).all(axis=1)]
197
+ if len(ww):
198
+ target=int(np.argmin(np.linalg.norm(grid-np.median(ww,axis=0),axis=1)))
199
+ target_delta,self.target_known=target_post_delta(self,grid,target)
200
+ self.last_target_delta=target_delta.copy()
201
+ if not dwell:self.offset=.92*self.offset+.08*target_delta
202
+ else:self.offset*=.85
203
+ is_release=t>0 and a[t,6]>.5 and a[t-1,6]<-.5
204
+ if is_release:self.last_release_t=t
205
+ if self.target_known and self.last_release_t is not None and 0<=t-self.last_release_t<8:
206
+ self.offset=self.last_target_delta.copy()
207
+ refws=ws[max(0,t-2):min(len(ws),t+3)]
208
+ validws=refws[np.isfinite(refws).all(axis=1)]
209
+ refw=np.median(validws,axis=0) if len(validws) else np.array([np.nan,np.nan])
210
+ w=wrist_hough(obs['rgb']['top_camera'])
211
+ if np.isfinite(w).all() and np.isfinite(refw).all():
212
+ error=refw+self.offset-w
213
+ if np.linalg.norm(refw-w)<35:
214
+ visual=np.clip(np.array([error[0],-error[1]])/75.,-0.06,0.06)
215
+ command[:2]=a[t,:2]+visual
216
+ visual_used=True
217
+ pause=False
218
+ if is_release and self.target_known and np.isfinite(w).all() and np.isfinite(refw).all():
219
+ error=refw+self.last_target_delta-w
220
+ count=self.release_wait.get(t,0)
221
+ if 3<np.linalg.norm(error)<35 and count<30 and (1000-step)>(len(a)-t):
222
+ command[:6]=0.;command[6]=-1.
223
+ command[:2]=np.clip(np.array([error[0],-error[1]])/75.,-0.06,0.06)
224
+ self.release_wait[t]=count+1;pause=True
225
+ if not pause:self.t+=1
226
+ self.last_diagnostic={'reason':'pose-aligned demonstration with bounded correction','demo':self.selected,'frame':t,'initial_mask_error':self.initial_score,'joint_error':float(np.linalg.norm(err)), 'visual_used':visual_used,'offset':self.offset.tolist(),'post_target_offset':target_delta.tolist(),'post_counts':self.post_counts.tolist(),'live_grid':None if self.live_grid is None else self.live_grid.tolist(),'release_pause':pause}
227
+ return np.clip(command,-1,1).astype(np.float32)
228
+ _controller=Controller()
229
+ def reset_episode():_controller.reset_episode()
230
+ def act(obs):return _controller.act(obs)
agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/demo_analysis.md ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Offline D20 demonstration analysis
2
+
3
+ Task: BuildTent-v1. Status: preparation only; no environment available, no exploration, no rehearsal, no success evaluation, no freeze. Exactly the 20 authorized entries in /demo_manifest.json were used. Only RGB, qpos, qvel, and actions were accessed through API.read_demo. No simulator, external task solution, hidden state, neural training, or network request was used.
4
+
5
+ ## Inspection ledger
6
+ Indices are zero-based observation indices. Actions at index t are treated as paired with observation t; observations have N+1 rows and actions N rows. This alignment follows the data layout, not a verified live transition test. Top and front images at every index below were actually viewed in evidence/sheet_0.jpg through sheet_3.jpg. Left and robot-camera images were actually viewed at the midpoint listed in evidence/side_0.jpg and side_1.jpg. Contact sheets preserve aspect ratios but reduce resolution. Numerical action/proprioception analysis covers every row of every trajectory. Automated RGB decoding/descriptor extraction is distinct from visual inspection and is recorded separately in candidate/memory_provenance.json.
7
+
8
+ | Demo | N | Viewed top/front frames | Viewed left/robot frame | Grip counts (-1/0/+1) | All grip changes | Last arm motion |
9
+ |---|---:|---|---:|---|---:|---:|
10
+ | demo_00 | 559 | 0, 139, 279, 419, 559 | 279 | 183/258/118 | 14 | 528 |
11
+ | demo_01 | 956 | 0, 239, 478, 717, 956 | 478 | 415/434/107 | 17 | 925 |
12
+ | demo_02 | 616 | 0, 154, 308, 462, 616 | 308 | 186/338/92 | 14 | 585 |
13
+ | demo_03 | 400 | 0, 100, 200, 300, 400 | 200 | 101/206/93 | 8 | 369 |
14
+ | demo_04 | 653 | 0, 163, 326, 489, 653 | 326 | 206/354/93 | 16 | 622 |
15
+ | demo_05 | 475 | 0, 118, 237, 356, 475 | 237 | 141/242/92 | 12 | 444 |
16
+ | demo_06 | 388 | 0, 97, 194, 291, 388 | 194 | 93/203/92 | 8 | 357 |
17
+ | demo_07 | 1000 | 0, 250, 500, 750, 1000 | 500 | 317/594/89 | 23 | 985 |
18
+ | demo_08 | 573 | 0, 143, 286, 429, 573 | 286 | 237/231/105 | 11 | 542 |
19
+ | demo_09 | 718 | 0, 179, 359, 538, 718 | 359 | 223/402/93 | 18 | 687 |
20
+ | demo_10 | 474 | 0, 118, 237, 355, 474 | 237 | 153/229/92 | 12 | 443 |
21
+ | demo_11 | 903 | 0, 225, 451, 677, 903 | 451 | 292/519/92 | 22 | 872 |
22
+ | demo_12 | 942 | 0, 235, 471, 706, 942 | 471 | 277/558/107 | 20 | 911 |
23
+ | demo_13 | 881 | 0, 220, 440, 660, 881 | 440 | 298/491/92 | 22 | 850 |
24
+ | demo_14 | 603 | 0, 150, 301, 452, 603 | 301 | 176/330/97 | 14 | 572 |
25
+ | demo_15 | 723 | 0, 180, 361, 542, 723 | 361 | 284/346/93 | 14 | 692 |
26
+ | demo_16 | 954 | 0, 238, 477, 715, 954 | 477 | 394/437/123 | 20 | 923 |
27
+ | demo_17 | 982 | 0, 245, 491, 736, 982 | 491 | 322/543/117 | 26 | 951 |
28
+ | demo_18 | 963 | 0, 240, 481, 722, 963 | 481 | 357/487/119 | 20 | 932 |
29
+ | demo_19 | 1000 | 0, 250, 500, 750, 1000 | 500 | 501/416/83 | 18 | 979 |
30
+
31
+ ## Per-trajectory visual evidence
32
+ - demo_00: Clockwise-tilted initial cloth; large fold at 279; final cloth flatter with corner rings visible. Completion is unverified.
33
+ - demo_01: Near-axis-aligned initial cloth; lifting and distortion at 717; final wrist occludes upper-right area. Completion is unverified.
34
+ - demo_02: Slight initial tilt; wrist traverses the far edge by 308; final cloth flatter, near square. Completion is unverified.
35
+ - demo_03: Opposite initial tilt to demo_00; corner manipulation at 100 and 300; final near-axis-aligned square. Completion is unverified.
36
+ - demo_04: Almost axis-aligned start; raised/folded corner at 163; final flatter square. Completion is unverified.
37
+ - demo_05: Clockwise initial tilt; hand over near-right corner at 118; final flatter square. Completion is unverified.
38
+ - demo_06: Clockwise initial tilt; corner deformation at 194; final flatter square. Completion is unverified.
39
+ - demo_07: Large rotation to diamond-like pose at 500; wavy cloth at 750; final arm occludes far-right area. Completion is unverified.
40
+ - demo_08: Near-axis-aligned start; lifted near edge at 143; final arm obscures lower edge. Completion is unverified.
41
+ - demo_09: Counterclockwise initial tilt; far-side manipulation at 359; final flatter square. Completion is unverified.
42
+ - demo_10: Clockwise initial tilt; crumpling near gripper in midpoint wrist view; final flatter cloth. Completion is unverified.
43
+ - demo_11: Counterclockwise initial tilt; far-right manipulation at 677; final lower-right area occluded. Completion is unverified.
44
+ - demo_12: Clockwise initial tilt; repeated different-corner poses in sampled frames; final upper edge occluded. Completion is unverified.
45
+ - demo_13: Counterclockwise initial tilt; near-corner manipulation at 440 and far-right at 660; final lower-right occluded. Completion is unverified.
46
+ - demo_14: Weak counterclockwise tilt; far-right handling at 150; final flatter near-axis-aligned cloth. Completion is unverified.
47
+ - demo_15: Counterclockwise initial tilt; far edge handled at 361; final flatter square. Completion is unverified.
48
+ - demo_16: Near-axis-aligned start; multiple corners handled across samples; lower-right deformation at 715; final flatter cloth. Completion is unverified.
49
+ - demo_17: Clockwise initial tilt; large cloth rotation at 736; final cloth flatter. Completion is unverified.
50
+ - demo_18: Counterclockwise initial tilt; strong folds at 481 and 722; final upper-right occluded. Completion is unverified.
51
+ - demo_19: Near-axis-aligned initial cloth; folds at 500; final lower-right corner still visibly lifted/occluded. Completion is unverified.
52
+
53
+ ## Action sequence and coordinate evidence
54
+ - There are 14,763 actions, with lengths 388–1,000 (19.4–50 seconds at the contracted 20 Hz). Demos 07 and 19 reach 1,000 actions; their length cannot establish failure or success.
55
+ - All trajectories begin with 25 stationary open-target actions; first arm movement is action 25. The initial qpos is the same nominal Panda pose in the inspected numeric data. This is robot initialization, not a scene identifier.
56
+ - Translation minima/maxima are [-0.08,-0.08,-0.08]/[0.08,0.08,0.08]. Rotation minima are approximately [-0.025450,-0.050815,-0.12], maxima [0.021432,0.007939,0.12]. Mean absolute components are [0.024094,0.023107,0.027012,0.001029,0.002315,0.019763,0.483980]. No label exceeds the contracted action bounds.
57
+ - Gripper labels are exactly -1, 0, +1. Under the contract these mean target mimic positions -0.01, 0.015, 0.04 meters; zero is NOT a preserve-current command. Qpos fingers need not reach a negative target under contact. Full transition indices and per-component extrema are in evidence/demo_stats.json. The gripper_switches field in that file counts changes in positivity only; all_gripper_transitions counts all three-valued changes and is the basis of the table.
58
+ - The label sequences contain 8–26 grip changes. Variable-duration intervening arm motion and repeated corner manipulation are consistent with approach, grasp, transport, lower, release, and reposition phases. Exact contact phase boundaries and retry reasons cannot be established from the sparse viewed frames.
59
+ - Most trajectories have 30 final stationary arm commands; demos 07 and 19 have shorter terminal stationary suffixes. Padding or settling must not be interpreted as an explicit success label.
60
+ - Contract order is one robot part of width seven: root-frame dx,dy,dz, three rotation components, absolute normalized Panda gripper target. Translation normalization maps ±1 to ±0.1 m, so observed translation labels correspond to at most 0.008 m requested deltas per call. These are requested deltas, not measured displacement.
61
+ - Rotation uses norm clipping then multiplication by -0.5 radians and Euler-to-quaternion conversion in a root-aligned frame. A positive normalized rotation component does not mean positive physical Euler angle under this implementation. Small roll/pitch labels and larger third-axis labels suggest primarily yaw adjustment; exact achieved orientation is not measured here.
62
+ - RGB coordinates are columns rightward and rows downward. Front/left/robot images are 240x320; top is 320x240. No camera calibration, depth, end-effector pose, post centers, cloth vertices, or grasp flags are available. Pixel-to-root metric coordinates are unknown; the draft does not invent them.
63
+
64
+ ## Observable completion and variation
65
+ All 20 starts show a pale square cloth with four dark corner openings, at varying planar angles and positions relative to the visible robot/table. Mid-trajectory cloth folds, shadows, arm occlusion, and large rotations vary substantially. Final images generally show a flatter, more axis-aligned cloth and ring-like corner features, but small posts can be hidden or coincide in projection. No success flags are present. It is not possible to certify all four threaded eyelets from these contact sheets; demo_19 particularly retains a raised/occluded near-right corner. Eyelet appearance in wrist views ranges from open holes to a light post-like structure within an opening (e.g. demo_06 frame 194), but that is local evidence only.
66
+
67
+ ## Draft and unresolved risks
68
+ The candidate uses a nonparametric memory of authorized observations and actions from all 20 demos. Fixed OpenCV RGB/occupancy descriptors and current qpos/qvel select compatible examples afresh each call. It never selects a demo by seed, assumes its scene is live, or advances a demonstration clock. Continuous commands are averaged only among three nearby samples with the same grip label and non-opposing translation. Gripper targets remain discrete. Sample selection retains every second action from 25 onward plus neighborhoods of every grip transition. Numeric scaling is a dataset statistic, not neural weight training.
69
+ White masks confuse robot and cloth; coarse images can erase eyelets; qpos preselection can exclude a useful visual neighbor; visually similar states may require different actions. Nearest-neighbor discontinuities, long stationary suffixes, contact hysteresis, and distribution shift may cause stalls or wrong grasps. The policy has no reliable semantic completion detector, force feedback, metric geometry, or validated recovery. It returns None only for exhausted native horizon or invalid proprioception; these are not success claims. Out-of-coverage inputs cause a bounded position-preserving gripper hold. The initial hold is entirely runner-owned. Native 20 Hz and 1,000-step horizon are unchanged; step_id 1 through 999 can produce actions after the charged hold.
70
+
71
+ ## Tools and validation limits
72
+ Used local Python, API.read_demo, NumPy, h5py (through API), OpenCV, shell commands, and local image viewing. The standalone API was imported via importlib. No RPC methods were called. No Internet/provider request was made; proxy availability was not tested. No simulator or live environment is available at this stage. Only syntax and synthetic-input checks are permitted and performed; see candidate/synthetic_checks.json. This is an unrehearsed draft, not a successful or frozen controller.
73
+
74
+ Validation outcome: all ten syntax/interface/synthetic checks passed. The initial hold assertion exceeded its default tolerance by a float32 rounding difference of 1.19e-7; the test tolerance was changed to 2e-7 and the checks passed on rerun. No controller change was required for that issue. The memory contains 7,900 examples. No action was sent to an environment.
75
+
76
+ ## Authorized API continuation
77
+ The preceding preparation record is preserved as historical evidence. Live continuation began with API.budget_status reporting 0/20 episodes, then /task_contract.json and /protocol_for_worker.json were read. Interactive/rehearsal evidence and exact candidate snapshots are under exploration/ and versions/.
78
+
79
+ Additional actual image inspections: demo_06 top frame 25 (saved as exploration/demo06_start.jpg), demo_06 top frame 55 (exploration/d55.jpg), and demo_16 top frame 100 (exploration/demo16_100.jpg). The latter shows a true wrist around (175,118) at demonstration resolution while an automated detector sometimes selected an arm joint around (78,142). This establishes a concrete localization failure, not task success.
80
+
81
+ Additional automated reads: all authorized qpos/qvel/actions to sequences.npz; all 20 authorized top images at frame 25 for initial silhouette choice; all authorized top frames in [0,N-1] for fixed contour/Hough features. Exact episode ranges are in sequences_provenance.json and geometry_provenance.json. geometry_clean.npz applies a fixed local-median rejection (>12 pixels) and interpolation only for gaps <=15 frames to these same features. No new neural weights, other-task data, benchmark implementation, simulator states, or external data were accessed.
82
+
83
+ Further live-stage analysis also visually inspected demo_16 top frame 150 (exploration/demo16_150.jpg). Post-cap extraction re-read only the same authorized top-camera ranges, processed every fifth frame from 25 onward, and retained exact processed-frame lists and persistent clusters in posts_inferred_provenance.json. The reference square is inferred from the two strongest adjacent cap clusters and oriented toward the initial cloth centroid; it is not a simulator-state read. Independent third/fourth visible clusters in several demos provide partial geometric checks, while weak inconsistent clusters remain recognized detector failures.
agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/demo_manifest.json ADDED
@@ -0,0 +1,1488 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task_id": "BuildTent-v1",
3
+ "status": "released",
4
+ "selection_rule": "prefer existing pinned D20, otherwise numeric trajectory ID sort before reading data",
5
+ "lineage_status": "pinned_comparator",
6
+ "episodes": [
7
+ {
8
+ "demo_id": "demo_00",
9
+ "file": "demo_00.h5",
10
+ "length": 559,
11
+ "fields": [
12
+ {
13
+ "path": "actions",
14
+ "shape": [
15
+ 559,
16
+ 7
17
+ ],
18
+ "dtype": "float32"
19
+ },
20
+ {
21
+ "path": "obs/agent/qpos",
22
+ "shape": [
23
+ 560,
24
+ 9
25
+ ],
26
+ "dtype": "float32"
27
+ },
28
+ {
29
+ "path": "obs/agent/qvel",
30
+ "shape": [
31
+ 560,
32
+ 9
33
+ ],
34
+ "dtype": "float32"
35
+ },
36
+ {
37
+ "path": "obs/sensor_data/front_camera/rgb",
38
+ "shape": [
39
+ 560,
40
+ 240,
41
+ 320,
42
+ 3
43
+ ],
44
+ "dtype": "uint8"
45
+ },
46
+ {
47
+ "path": "obs/sensor_data/left_camera/rgb",
48
+ "shape": [
49
+ 560,
50
+ 240,
51
+ 320,
52
+ 3
53
+ ],
54
+ "dtype": "uint8"
55
+ },
56
+ {
57
+ "path": "obs/sensor_data/robot_camera/rgb",
58
+ "shape": [
59
+ 560,
60
+ 240,
61
+ 320,
62
+ 3
63
+ ],
64
+ "dtype": "uint8"
65
+ },
66
+ {
67
+ "path": "obs/sensor_data/top_camera/rgb",
68
+ "shape": [
69
+ 560,
70
+ 320,
71
+ 240,
72
+ 3
73
+ ],
74
+ "dtype": "uint8"
75
+ }
76
+ ],
77
+ "trajectory_sha256": "7c70d76b07175afc25d46cf29607e43d28459b1942afa8bf63e60c7f993565f4",
78
+ "source_sha256": "3adebfc271ac799b4b3ad34c43d83ede73cf9a3f4e7236a622629fd2f936d882",
79
+ "view_sha256": "15c3b69574899adcdf36f92fa49659d450c82d41ba04430ef4bc92f3f97f7a38"
80
+ },
81
+ {
82
+ "demo_id": "demo_01",
83
+ "file": "demo_01.h5",
84
+ "length": 956,
85
+ "fields": [
86
+ {
87
+ "path": "actions",
88
+ "shape": [
89
+ 956,
90
+ 7
91
+ ],
92
+ "dtype": "float32"
93
+ },
94
+ {
95
+ "path": "obs/agent/qpos",
96
+ "shape": [
97
+ 957,
98
+ 9
99
+ ],
100
+ "dtype": "float32"
101
+ },
102
+ {
103
+ "path": "obs/agent/qvel",
104
+ "shape": [
105
+ 957,
106
+ 9
107
+ ],
108
+ "dtype": "float32"
109
+ },
110
+ {
111
+ "path": "obs/sensor_data/front_camera/rgb",
112
+ "shape": [
113
+ 957,
114
+ 240,
115
+ 320,
116
+ 3
117
+ ],
118
+ "dtype": "uint8"
119
+ },
120
+ {
121
+ "path": "obs/sensor_data/left_camera/rgb",
122
+ "shape": [
123
+ 957,
124
+ 240,
125
+ 320,
126
+ 3
127
+ ],
128
+ "dtype": "uint8"
129
+ },
130
+ {
131
+ "path": "obs/sensor_data/robot_camera/rgb",
132
+ "shape": [
133
+ 957,
134
+ 240,
135
+ 320,
136
+ 3
137
+ ],
138
+ "dtype": "uint8"
139
+ },
140
+ {
141
+ "path": "obs/sensor_data/top_camera/rgb",
142
+ "shape": [
143
+ 957,
144
+ 320,
145
+ 240,
146
+ 3
147
+ ],
148
+ "dtype": "uint8"
149
+ }
150
+ ],
151
+ "trajectory_sha256": "a1ef70e5976bb0754184923dade866a3625224bbe029f78603e31d3d091e0268",
152
+ "source_sha256": "198d92fa64b53924ef54eccacf56a3d6acee16096f87b20e29712bf7a5e0a2a6",
153
+ "view_sha256": "ff58ffc7b142bc0b1bf201f99b2f48ea9e7392dec171330cf67ee3f4b3296d17"
154
+ },
155
+ {
156
+ "demo_id": "demo_02",
157
+ "file": "demo_02.h5",
158
+ "length": 616,
159
+ "fields": [
160
+ {
161
+ "path": "actions",
162
+ "shape": [
163
+ 616,
164
+ 7
165
+ ],
166
+ "dtype": "float32"
167
+ },
168
+ {
169
+ "path": "obs/agent/qpos",
170
+ "shape": [
171
+ 617,
172
+ 9
173
+ ],
174
+ "dtype": "float32"
175
+ },
176
+ {
177
+ "path": "obs/agent/qvel",
178
+ "shape": [
179
+ 617,
180
+ 9
181
+ ],
182
+ "dtype": "float32"
183
+ },
184
+ {
185
+ "path": "obs/sensor_data/front_camera/rgb",
186
+ "shape": [
187
+ 617,
188
+ 240,
189
+ 320,
190
+ 3
191
+ ],
192
+ "dtype": "uint8"
193
+ },
194
+ {
195
+ "path": "obs/sensor_data/left_camera/rgb",
196
+ "shape": [
197
+ 617,
198
+ 240,
199
+ 320,
200
+ 3
201
+ ],
202
+ "dtype": "uint8"
203
+ },
204
+ {
205
+ "path": "obs/sensor_data/robot_camera/rgb",
206
+ "shape": [
207
+ 617,
208
+ 240,
209
+ 320,
210
+ 3
211
+ ],
212
+ "dtype": "uint8"
213
+ },
214
+ {
215
+ "path": "obs/sensor_data/top_camera/rgb",
216
+ "shape": [
217
+ 617,
218
+ 320,
219
+ 240,
220
+ 3
221
+ ],
222
+ "dtype": "uint8"
223
+ }
224
+ ],
225
+ "trajectory_sha256": "94a98c2d244317f0826c189f1a6b6aa9413c7217b3917f67a88613d37bc179b0",
226
+ "source_sha256": "a6d04985efff1114fc1d4a58cb07dfd5999d577f612c3396d14ee3b71b8b1d13",
227
+ "view_sha256": "dc859b2fe4da9ed074a2a4e408828ec07e5510974096958bc51b75b133765cea"
228
+ },
229
+ {
230
+ "demo_id": "demo_03",
231
+ "file": "demo_03.h5",
232
+ "length": 400,
233
+ "fields": [
234
+ {
235
+ "path": "actions",
236
+ "shape": [
237
+ 400,
238
+ 7
239
+ ],
240
+ "dtype": "float32"
241
+ },
242
+ {
243
+ "path": "obs/agent/qpos",
244
+ "shape": [
245
+ 401,
246
+ 9
247
+ ],
248
+ "dtype": "float32"
249
+ },
250
+ {
251
+ "path": "obs/agent/qvel",
252
+ "shape": [
253
+ 401,
254
+ 9
255
+ ],
256
+ "dtype": "float32"
257
+ },
258
+ {
259
+ "path": "obs/sensor_data/front_camera/rgb",
260
+ "shape": [
261
+ 401,
262
+ 240,
263
+ 320,
264
+ 3
265
+ ],
266
+ "dtype": "uint8"
267
+ },
268
+ {
269
+ "path": "obs/sensor_data/left_camera/rgb",
270
+ "shape": [
271
+ 401,
272
+ 240,
273
+ 320,
274
+ 3
275
+ ],
276
+ "dtype": "uint8"
277
+ },
278
+ {
279
+ "path": "obs/sensor_data/robot_camera/rgb",
280
+ "shape": [
281
+ 401,
282
+ 240,
283
+ 320,
284
+ 3
285
+ ],
286
+ "dtype": "uint8"
287
+ },
288
+ {
289
+ "path": "obs/sensor_data/top_camera/rgb",
290
+ "shape": [
291
+ 401,
292
+ 320,
293
+ 240,
294
+ 3
295
+ ],
296
+ "dtype": "uint8"
297
+ }
298
+ ],
299
+ "trajectory_sha256": "e425ec4917851f9cd2211afd1ac030a6ee61f6e9837485d1ed47293cf2b203a6",
300
+ "source_sha256": "9332c295948253c08f55e0d644f174cbde398303a2fd0ebaf9193376b6d9d7ee",
301
+ "view_sha256": "717c1cc66f0fddceb7c3a4b73cab9234c9ac9ebcd10cb0b390ba19c314d67bcd"
302
+ },
303
+ {
304
+ "demo_id": "demo_04",
305
+ "file": "demo_04.h5",
306
+ "length": 653,
307
+ "fields": [
308
+ {
309
+ "path": "actions",
310
+ "shape": [
311
+ 653,
312
+ 7
313
+ ],
314
+ "dtype": "float32"
315
+ },
316
+ {
317
+ "path": "obs/agent/qpos",
318
+ "shape": [
319
+ 654,
320
+ 9
321
+ ],
322
+ "dtype": "float32"
323
+ },
324
+ {
325
+ "path": "obs/agent/qvel",
326
+ "shape": [
327
+ 654,
328
+ 9
329
+ ],
330
+ "dtype": "float32"
331
+ },
332
+ {
333
+ "path": "obs/sensor_data/front_camera/rgb",
334
+ "shape": [
335
+ 654,
336
+ 240,
337
+ 320,
338
+ 3
339
+ ],
340
+ "dtype": "uint8"
341
+ },
342
+ {
343
+ "path": "obs/sensor_data/left_camera/rgb",
344
+ "shape": [
345
+ 654,
346
+ 240,
347
+ 320,
348
+ 3
349
+ ],
350
+ "dtype": "uint8"
351
+ },
352
+ {
353
+ "path": "obs/sensor_data/robot_camera/rgb",
354
+ "shape": [
355
+ 654,
356
+ 240,
357
+ 320,
358
+ 3
359
+ ],
360
+ "dtype": "uint8"
361
+ },
362
+ {
363
+ "path": "obs/sensor_data/top_camera/rgb",
364
+ "shape": [
365
+ 654,
366
+ 320,
367
+ 240,
368
+ 3
369
+ ],
370
+ "dtype": "uint8"
371
+ }
372
+ ],
373
+ "trajectory_sha256": "99ce80ff58dd56320e471d5f5ed32278f82e169518f04a90c7052c1c3e7c3de3",
374
+ "source_sha256": "946cee8fa736ae3ecf8b337a977e4177cdc6227dfa672fd9a82e51cedd08e236",
375
+ "view_sha256": "850ac65eaa41c4f8c1f558131a04116de2fc6c618263facfb33c136f5f44e251"
376
+ },
377
+ {
378
+ "demo_id": "demo_05",
379
+ "file": "demo_05.h5",
380
+ "length": 475,
381
+ "fields": [
382
+ {
383
+ "path": "actions",
384
+ "shape": [
385
+ 475,
386
+ 7
387
+ ],
388
+ "dtype": "float32"
389
+ },
390
+ {
391
+ "path": "obs/agent/qpos",
392
+ "shape": [
393
+ 476,
394
+ 9
395
+ ],
396
+ "dtype": "float32"
397
+ },
398
+ {
399
+ "path": "obs/agent/qvel",
400
+ "shape": [
401
+ 476,
402
+ 9
403
+ ],
404
+ "dtype": "float32"
405
+ },
406
+ {
407
+ "path": "obs/sensor_data/front_camera/rgb",
408
+ "shape": [
409
+ 476,
410
+ 240,
411
+ 320,
412
+ 3
413
+ ],
414
+ "dtype": "uint8"
415
+ },
416
+ {
417
+ "path": "obs/sensor_data/left_camera/rgb",
418
+ "shape": [
419
+ 476,
420
+ 240,
421
+ 320,
422
+ 3
423
+ ],
424
+ "dtype": "uint8"
425
+ },
426
+ {
427
+ "path": "obs/sensor_data/robot_camera/rgb",
428
+ "shape": [
429
+ 476,
430
+ 240,
431
+ 320,
432
+ 3
433
+ ],
434
+ "dtype": "uint8"
435
+ },
436
+ {
437
+ "path": "obs/sensor_data/top_camera/rgb",
438
+ "shape": [
439
+ 476,
440
+ 320,
441
+ 240,
442
+ 3
443
+ ],
444
+ "dtype": "uint8"
445
+ }
446
+ ],
447
+ "trajectory_sha256": "d9fd1c3d54c877a673fbf1e6d04bff8a50cf36826fc97e07981693178edea334",
448
+ "source_sha256": "d3771d0c2079b1d4f68f06d0e86a75af13ef98703d45c6440789d12bc6e6800e",
449
+ "view_sha256": "f8a6bce620a462745d908de9caa0e2a9a4e66af43c3cfd6a076ac139fd8eb1d6"
450
+ },
451
+ {
452
+ "demo_id": "demo_06",
453
+ "file": "demo_06.h5",
454
+ "length": 388,
455
+ "fields": [
456
+ {
457
+ "path": "actions",
458
+ "shape": [
459
+ 388,
460
+ 7
461
+ ],
462
+ "dtype": "float32"
463
+ },
464
+ {
465
+ "path": "obs/agent/qpos",
466
+ "shape": [
467
+ 389,
468
+ 9
469
+ ],
470
+ "dtype": "float32"
471
+ },
472
+ {
473
+ "path": "obs/agent/qvel",
474
+ "shape": [
475
+ 389,
476
+ 9
477
+ ],
478
+ "dtype": "float32"
479
+ },
480
+ {
481
+ "path": "obs/sensor_data/front_camera/rgb",
482
+ "shape": [
483
+ 389,
484
+ 240,
485
+ 320,
486
+ 3
487
+ ],
488
+ "dtype": "uint8"
489
+ },
490
+ {
491
+ "path": "obs/sensor_data/left_camera/rgb",
492
+ "shape": [
493
+ 389,
494
+ 240,
495
+ 320,
496
+ 3
497
+ ],
498
+ "dtype": "uint8"
499
+ },
500
+ {
501
+ "path": "obs/sensor_data/robot_camera/rgb",
502
+ "shape": [
503
+ 389,
504
+ 240,
505
+ 320,
506
+ 3
507
+ ],
508
+ "dtype": "uint8"
509
+ },
510
+ {
511
+ "path": "obs/sensor_data/top_camera/rgb",
512
+ "shape": [
513
+ 389,
514
+ 320,
515
+ 240,
516
+ 3
517
+ ],
518
+ "dtype": "uint8"
519
+ }
520
+ ],
521
+ "trajectory_sha256": "776f6d74524907fefdd7f2eef34ea83672d24d2c5ed6687e369582dad40c94e0",
522
+ "source_sha256": "01f472e0e88464242dd9c5f5179cbc5fc8290e83d04ebd1c8ebb03121be7c61d",
523
+ "view_sha256": "72fecef57643178c586ebe1bb3723f933837d795e17fb8b208e18cddc2308b3d"
524
+ },
525
+ {
526
+ "demo_id": "demo_07",
527
+ "file": "demo_07.h5",
528
+ "length": 1000,
529
+ "fields": [
530
+ {
531
+ "path": "actions",
532
+ "shape": [
533
+ 1000,
534
+ 7
535
+ ],
536
+ "dtype": "float32"
537
+ },
538
+ {
539
+ "path": "obs/agent/qpos",
540
+ "shape": [
541
+ 1001,
542
+ 9
543
+ ],
544
+ "dtype": "float32"
545
+ },
546
+ {
547
+ "path": "obs/agent/qvel",
548
+ "shape": [
549
+ 1001,
550
+ 9
551
+ ],
552
+ "dtype": "float32"
553
+ },
554
+ {
555
+ "path": "obs/sensor_data/front_camera/rgb",
556
+ "shape": [
557
+ 1001,
558
+ 240,
559
+ 320,
560
+ 3
561
+ ],
562
+ "dtype": "uint8"
563
+ },
564
+ {
565
+ "path": "obs/sensor_data/left_camera/rgb",
566
+ "shape": [
567
+ 1001,
568
+ 240,
569
+ 320,
570
+ 3
571
+ ],
572
+ "dtype": "uint8"
573
+ },
574
+ {
575
+ "path": "obs/sensor_data/robot_camera/rgb",
576
+ "shape": [
577
+ 1001,
578
+ 240,
579
+ 320,
580
+ 3
581
+ ],
582
+ "dtype": "uint8"
583
+ },
584
+ {
585
+ "path": "obs/sensor_data/top_camera/rgb",
586
+ "shape": [
587
+ 1001,
588
+ 320,
589
+ 240,
590
+ 3
591
+ ],
592
+ "dtype": "uint8"
593
+ }
594
+ ],
595
+ "trajectory_sha256": "a0f5f5ce15aba4102a9c6e6ab80eaee580989ddf079396ae192801ab2d20a565",
596
+ "source_sha256": "02c15e9c88d1364b01d57b6cd70a10a59ffdfad319809acf85865c05ee006078",
597
+ "view_sha256": "248cf45c67c4a7b039b64214d67aca7426df091b3ed888fb225a97a38ea84848"
598
+ },
599
+ {
600
+ "demo_id": "demo_08",
601
+ "file": "demo_08.h5",
602
+ "length": 573,
603
+ "fields": [
604
+ {
605
+ "path": "actions",
606
+ "shape": [
607
+ 573,
608
+ 7
609
+ ],
610
+ "dtype": "float32"
611
+ },
612
+ {
613
+ "path": "obs/agent/qpos",
614
+ "shape": [
615
+ 574,
616
+ 9
617
+ ],
618
+ "dtype": "float32"
619
+ },
620
+ {
621
+ "path": "obs/agent/qvel",
622
+ "shape": [
623
+ 574,
624
+ 9
625
+ ],
626
+ "dtype": "float32"
627
+ },
628
+ {
629
+ "path": "obs/sensor_data/front_camera/rgb",
630
+ "shape": [
631
+ 574,
632
+ 240,
633
+ 320,
634
+ 3
635
+ ],
636
+ "dtype": "uint8"
637
+ },
638
+ {
639
+ "path": "obs/sensor_data/left_camera/rgb",
640
+ "shape": [
641
+ 574,
642
+ 240,
643
+ 320,
644
+ 3
645
+ ],
646
+ "dtype": "uint8"
647
+ },
648
+ {
649
+ "path": "obs/sensor_data/robot_camera/rgb",
650
+ "shape": [
651
+ 574,
652
+ 240,
653
+ 320,
654
+ 3
655
+ ],
656
+ "dtype": "uint8"
657
+ },
658
+ {
659
+ "path": "obs/sensor_data/top_camera/rgb",
660
+ "shape": [
661
+ 574,
662
+ 320,
663
+ 240,
664
+ 3
665
+ ],
666
+ "dtype": "uint8"
667
+ }
668
+ ],
669
+ "trajectory_sha256": "d64cb28ef90c755b2410b43f96317f6ada77ac1f51117de155153ed7ca8e643d",
670
+ "source_sha256": "ea5a0a039f85cc5f41b8fab909380d0265317755e2ecbecbbb9d3ca77a93acd6",
671
+ "view_sha256": "cf3747c750a57ae620d89325dd49dbe5810c861f0f1a4fae22b50f48bc79437f"
672
+ },
673
+ {
674
+ "demo_id": "demo_09",
675
+ "file": "demo_09.h5",
676
+ "length": 718,
677
+ "fields": [
678
+ {
679
+ "path": "actions",
680
+ "shape": [
681
+ 718,
682
+ 7
683
+ ],
684
+ "dtype": "float32"
685
+ },
686
+ {
687
+ "path": "obs/agent/qpos",
688
+ "shape": [
689
+ 719,
690
+ 9
691
+ ],
692
+ "dtype": "float32"
693
+ },
694
+ {
695
+ "path": "obs/agent/qvel",
696
+ "shape": [
697
+ 719,
698
+ 9
699
+ ],
700
+ "dtype": "float32"
701
+ },
702
+ {
703
+ "path": "obs/sensor_data/front_camera/rgb",
704
+ "shape": [
705
+ 719,
706
+ 240,
707
+ 320,
708
+ 3
709
+ ],
710
+ "dtype": "uint8"
711
+ },
712
+ {
713
+ "path": "obs/sensor_data/left_camera/rgb",
714
+ "shape": [
715
+ 719,
716
+ 240,
717
+ 320,
718
+ 3
719
+ ],
720
+ "dtype": "uint8"
721
+ },
722
+ {
723
+ "path": "obs/sensor_data/robot_camera/rgb",
724
+ "shape": [
725
+ 719,
726
+ 240,
727
+ 320,
728
+ 3
729
+ ],
730
+ "dtype": "uint8"
731
+ },
732
+ {
733
+ "path": "obs/sensor_data/top_camera/rgb",
734
+ "shape": [
735
+ 719,
736
+ 320,
737
+ 240,
738
+ 3
739
+ ],
740
+ "dtype": "uint8"
741
+ }
742
+ ],
743
+ "trajectory_sha256": "f364a4419f2549a167de086a73b81c70dd4eb063b52927bb2555748c0c881caf",
744
+ "source_sha256": "5c561f5cd6b71fb340ad808b7db29e7f561ceb6ff7f87e6fe6c6d6c5f0ea2a53",
745
+ "view_sha256": "562f8d82bcfb0e30d70379048e776a83cb7602cd5129fb31e5cc3de987055cf6"
746
+ },
747
+ {
748
+ "demo_id": "demo_10",
749
+ "file": "demo_10.h5",
750
+ "length": 474,
751
+ "fields": [
752
+ {
753
+ "path": "actions",
754
+ "shape": [
755
+ 474,
756
+ 7
757
+ ],
758
+ "dtype": "float32"
759
+ },
760
+ {
761
+ "path": "obs/agent/qpos",
762
+ "shape": [
763
+ 475,
764
+ 9
765
+ ],
766
+ "dtype": "float32"
767
+ },
768
+ {
769
+ "path": "obs/agent/qvel",
770
+ "shape": [
771
+ 475,
772
+ 9
773
+ ],
774
+ "dtype": "float32"
775
+ },
776
+ {
777
+ "path": "obs/sensor_data/front_camera/rgb",
778
+ "shape": [
779
+ 475,
780
+ 240,
781
+ 320,
782
+ 3
783
+ ],
784
+ "dtype": "uint8"
785
+ },
786
+ {
787
+ "path": "obs/sensor_data/left_camera/rgb",
788
+ "shape": [
789
+ 475,
790
+ 240,
791
+ 320,
792
+ 3
793
+ ],
794
+ "dtype": "uint8"
795
+ },
796
+ {
797
+ "path": "obs/sensor_data/robot_camera/rgb",
798
+ "shape": [
799
+ 475,
800
+ 240,
801
+ 320,
802
+ 3
803
+ ],
804
+ "dtype": "uint8"
805
+ },
806
+ {
807
+ "path": "obs/sensor_data/top_camera/rgb",
808
+ "shape": [
809
+ 475,
810
+ 320,
811
+ 240,
812
+ 3
813
+ ],
814
+ "dtype": "uint8"
815
+ }
816
+ ],
817
+ "trajectory_sha256": "bda95b621788ca1c24bf96dab8154eaa1f8901b22d3ceb6c6c4bc970ded6f89d",
818
+ "source_sha256": "d5fefb5fc16ad9f68d61ee9e987be31f27a3678e12ce912dc522da7993e2d65f",
819
+ "view_sha256": "2a340114bd0bc728d57b0e734774f5b16f5278e7e4a2ac4880188e817a6606f6"
820
+ },
821
+ {
822
+ "demo_id": "demo_11",
823
+ "file": "demo_11.h5",
824
+ "length": 903,
825
+ "fields": [
826
+ {
827
+ "path": "actions",
828
+ "shape": [
829
+ 903,
830
+ 7
831
+ ],
832
+ "dtype": "float32"
833
+ },
834
+ {
835
+ "path": "obs/agent/qpos",
836
+ "shape": [
837
+ 904,
838
+ 9
839
+ ],
840
+ "dtype": "float32"
841
+ },
842
+ {
843
+ "path": "obs/agent/qvel",
844
+ "shape": [
845
+ 904,
846
+ 9
847
+ ],
848
+ "dtype": "float32"
849
+ },
850
+ {
851
+ "path": "obs/sensor_data/front_camera/rgb",
852
+ "shape": [
853
+ 904,
854
+ 240,
855
+ 320,
856
+ 3
857
+ ],
858
+ "dtype": "uint8"
859
+ },
860
+ {
861
+ "path": "obs/sensor_data/left_camera/rgb",
862
+ "shape": [
863
+ 904,
864
+ 240,
865
+ 320,
866
+ 3
867
+ ],
868
+ "dtype": "uint8"
869
+ },
870
+ {
871
+ "path": "obs/sensor_data/robot_camera/rgb",
872
+ "shape": [
873
+ 904,
874
+ 240,
875
+ 320,
876
+ 3
877
+ ],
878
+ "dtype": "uint8"
879
+ },
880
+ {
881
+ "path": "obs/sensor_data/top_camera/rgb",
882
+ "shape": [
883
+ 904,
884
+ 320,
885
+ 240,
886
+ 3
887
+ ],
888
+ "dtype": "uint8"
889
+ }
890
+ ],
891
+ "trajectory_sha256": "c190403370cc1efd2b5b3655c0597592ff800d02b80ec854c7d5b12a07869c47",
892
+ "source_sha256": "e4146a72f1a5d03d470ed94d9aed2a27899353dc07e3ccb2e416771682a6d8f9",
893
+ "view_sha256": "3e3675b75fe4ed84fbb7e99a2cb94ec133d6b89ada697d350b3fd71fe6b4c527"
894
+ },
895
+ {
896
+ "demo_id": "demo_12",
897
+ "file": "demo_12.h5",
898
+ "length": 942,
899
+ "fields": [
900
+ {
901
+ "path": "actions",
902
+ "shape": [
903
+ 942,
904
+ 7
905
+ ],
906
+ "dtype": "float32"
907
+ },
908
+ {
909
+ "path": "obs/agent/qpos",
910
+ "shape": [
911
+ 943,
912
+ 9
913
+ ],
914
+ "dtype": "float32"
915
+ },
916
+ {
917
+ "path": "obs/agent/qvel",
918
+ "shape": [
919
+ 943,
920
+ 9
921
+ ],
922
+ "dtype": "float32"
923
+ },
924
+ {
925
+ "path": "obs/sensor_data/front_camera/rgb",
926
+ "shape": [
927
+ 943,
928
+ 240,
929
+ 320,
930
+ 3
931
+ ],
932
+ "dtype": "uint8"
933
+ },
934
+ {
935
+ "path": "obs/sensor_data/left_camera/rgb",
936
+ "shape": [
937
+ 943,
938
+ 240,
939
+ 320,
940
+ 3
941
+ ],
942
+ "dtype": "uint8"
943
+ },
944
+ {
945
+ "path": "obs/sensor_data/robot_camera/rgb",
946
+ "shape": [
947
+ 943,
948
+ 240,
949
+ 320,
950
+ 3
951
+ ],
952
+ "dtype": "uint8"
953
+ },
954
+ {
955
+ "path": "obs/sensor_data/top_camera/rgb",
956
+ "shape": [
957
+ 943,
958
+ 320,
959
+ 240,
960
+ 3
961
+ ],
962
+ "dtype": "uint8"
963
+ }
964
+ ],
965
+ "trajectory_sha256": "6c306edf8d9a2b636b0a51cee36cc9953352f65b2dc9fb254f5f3c64bc25b53a",
966
+ "source_sha256": "3cfc20951b7f5fa06607c50e6f802aed62ea2138d4d0603f353ec335566a0020",
967
+ "view_sha256": "07d691b9b7b2fca308d5e116f37c0134ca2f81c94e9dba370475174105704ca4"
968
+ },
969
+ {
970
+ "demo_id": "demo_13",
971
+ "file": "demo_13.h5",
972
+ "length": 881,
973
+ "fields": [
974
+ {
975
+ "path": "actions",
976
+ "shape": [
977
+ 881,
978
+ 7
979
+ ],
980
+ "dtype": "float32"
981
+ },
982
+ {
983
+ "path": "obs/agent/qpos",
984
+ "shape": [
985
+ 882,
986
+ 9
987
+ ],
988
+ "dtype": "float32"
989
+ },
990
+ {
991
+ "path": "obs/agent/qvel",
992
+ "shape": [
993
+ 882,
994
+ 9
995
+ ],
996
+ "dtype": "float32"
997
+ },
998
+ {
999
+ "path": "obs/sensor_data/front_camera/rgb",
1000
+ "shape": [
1001
+ 882,
1002
+ 240,
1003
+ 320,
1004
+ 3
1005
+ ],
1006
+ "dtype": "uint8"
1007
+ },
1008
+ {
1009
+ "path": "obs/sensor_data/left_camera/rgb",
1010
+ "shape": [
1011
+ 882,
1012
+ 240,
1013
+ 320,
1014
+ 3
1015
+ ],
1016
+ "dtype": "uint8"
1017
+ },
1018
+ {
1019
+ "path": "obs/sensor_data/robot_camera/rgb",
1020
+ "shape": [
1021
+ 882,
1022
+ 240,
1023
+ 320,
1024
+ 3
1025
+ ],
1026
+ "dtype": "uint8"
1027
+ },
1028
+ {
1029
+ "path": "obs/sensor_data/top_camera/rgb",
1030
+ "shape": [
1031
+ 882,
1032
+ 320,
1033
+ 240,
1034
+ 3
1035
+ ],
1036
+ "dtype": "uint8"
1037
+ }
1038
+ ],
1039
+ "trajectory_sha256": "9ad364897afbe401b82f1c64cdd51cbac81bd79a4632e0c3f085ccd94e830a78",
1040
+ "source_sha256": "5136e72f02d9e10b5c88d2c2e4d525edb63be7ca49567d9fc27f8a7726e3bd30",
1041
+ "view_sha256": "fbe474c4ad0081428dcb84009f37cf60a93264b214860ceca0c88d99f8da6eb1"
1042
+ },
1043
+ {
1044
+ "demo_id": "demo_14",
1045
+ "file": "demo_14.h5",
1046
+ "length": 603,
1047
+ "fields": [
1048
+ {
1049
+ "path": "actions",
1050
+ "shape": [
1051
+ 603,
1052
+ 7
1053
+ ],
1054
+ "dtype": "float32"
1055
+ },
1056
+ {
1057
+ "path": "obs/agent/qpos",
1058
+ "shape": [
1059
+ 604,
1060
+ 9
1061
+ ],
1062
+ "dtype": "float32"
1063
+ },
1064
+ {
1065
+ "path": "obs/agent/qvel",
1066
+ "shape": [
1067
+ 604,
1068
+ 9
1069
+ ],
1070
+ "dtype": "float32"
1071
+ },
1072
+ {
1073
+ "path": "obs/sensor_data/front_camera/rgb",
1074
+ "shape": [
1075
+ 604,
1076
+ 240,
1077
+ 320,
1078
+ 3
1079
+ ],
1080
+ "dtype": "uint8"
1081
+ },
1082
+ {
1083
+ "path": "obs/sensor_data/left_camera/rgb",
1084
+ "shape": [
1085
+ 604,
1086
+ 240,
1087
+ 320,
1088
+ 3
1089
+ ],
1090
+ "dtype": "uint8"
1091
+ },
1092
+ {
1093
+ "path": "obs/sensor_data/robot_camera/rgb",
1094
+ "shape": [
1095
+ 604,
1096
+ 240,
1097
+ 320,
1098
+ 3
1099
+ ],
1100
+ "dtype": "uint8"
1101
+ },
1102
+ {
1103
+ "path": "obs/sensor_data/top_camera/rgb",
1104
+ "shape": [
1105
+ 604,
1106
+ 320,
1107
+ 240,
1108
+ 3
1109
+ ],
1110
+ "dtype": "uint8"
1111
+ }
1112
+ ],
1113
+ "trajectory_sha256": "7eb57bd0241ea51bfa598372df16139655c546beb3c3af2bdd9e33db5ce37203",
1114
+ "source_sha256": "9ef83632be946ee48c1fa07be7521c277b1bc3828a35ed0dbc5c1ce1cd97b2c0",
1115
+ "view_sha256": "3f5c21bebcbc36bd635198c94a5c9621c7eaf46f1830392a6a12173c533fbef1"
1116
+ },
1117
+ {
1118
+ "demo_id": "demo_15",
1119
+ "file": "demo_15.h5",
1120
+ "length": 723,
1121
+ "fields": [
1122
+ {
1123
+ "path": "actions",
1124
+ "shape": [
1125
+ 723,
1126
+ 7
1127
+ ],
1128
+ "dtype": "float32"
1129
+ },
1130
+ {
1131
+ "path": "obs/agent/qpos",
1132
+ "shape": [
1133
+ 724,
1134
+ 9
1135
+ ],
1136
+ "dtype": "float32"
1137
+ },
1138
+ {
1139
+ "path": "obs/agent/qvel",
1140
+ "shape": [
1141
+ 724,
1142
+ 9
1143
+ ],
1144
+ "dtype": "float32"
1145
+ },
1146
+ {
1147
+ "path": "obs/sensor_data/front_camera/rgb",
1148
+ "shape": [
1149
+ 724,
1150
+ 240,
1151
+ 320,
1152
+ 3
1153
+ ],
1154
+ "dtype": "uint8"
1155
+ },
1156
+ {
1157
+ "path": "obs/sensor_data/left_camera/rgb",
1158
+ "shape": [
1159
+ 724,
1160
+ 240,
1161
+ 320,
1162
+ 3
1163
+ ],
1164
+ "dtype": "uint8"
1165
+ },
1166
+ {
1167
+ "path": "obs/sensor_data/robot_camera/rgb",
1168
+ "shape": [
1169
+ 724,
1170
+ 240,
1171
+ 320,
1172
+ 3
1173
+ ],
1174
+ "dtype": "uint8"
1175
+ },
1176
+ {
1177
+ "path": "obs/sensor_data/top_camera/rgb",
1178
+ "shape": [
1179
+ 724,
1180
+ 320,
1181
+ 240,
1182
+ 3
1183
+ ],
1184
+ "dtype": "uint8"
1185
+ }
1186
+ ],
1187
+ "trajectory_sha256": "da99ec7288518e899cf11b5bbfe198417d97299137d8c3efa191ef57f2595a2b",
1188
+ "source_sha256": "ed5defad07262760db5bf76624c18732754b1c2ff5deecb4adf6d63f4d1771da",
1189
+ "view_sha256": "2fea6081c183a51095ff8f5f548e21d50e9f4b163b4c3195ab8db0fb38cee857"
1190
+ },
1191
+ {
1192
+ "demo_id": "demo_16",
1193
+ "file": "demo_16.h5",
1194
+ "length": 954,
1195
+ "fields": [
1196
+ {
1197
+ "path": "actions",
1198
+ "shape": [
1199
+ 954,
1200
+ 7
1201
+ ],
1202
+ "dtype": "float32"
1203
+ },
1204
+ {
1205
+ "path": "obs/agent/qpos",
1206
+ "shape": [
1207
+ 955,
1208
+ 9
1209
+ ],
1210
+ "dtype": "float32"
1211
+ },
1212
+ {
1213
+ "path": "obs/agent/qvel",
1214
+ "shape": [
1215
+ 955,
1216
+ 9
1217
+ ],
1218
+ "dtype": "float32"
1219
+ },
1220
+ {
1221
+ "path": "obs/sensor_data/front_camera/rgb",
1222
+ "shape": [
1223
+ 955,
1224
+ 240,
1225
+ 320,
1226
+ 3
1227
+ ],
1228
+ "dtype": "uint8"
1229
+ },
1230
+ {
1231
+ "path": "obs/sensor_data/left_camera/rgb",
1232
+ "shape": [
1233
+ 955,
1234
+ 240,
1235
+ 320,
1236
+ 3
1237
+ ],
1238
+ "dtype": "uint8"
1239
+ },
1240
+ {
1241
+ "path": "obs/sensor_data/robot_camera/rgb",
1242
+ "shape": [
1243
+ 955,
1244
+ 240,
1245
+ 320,
1246
+ 3
1247
+ ],
1248
+ "dtype": "uint8"
1249
+ },
1250
+ {
1251
+ "path": "obs/sensor_data/top_camera/rgb",
1252
+ "shape": [
1253
+ 955,
1254
+ 320,
1255
+ 240,
1256
+ 3
1257
+ ],
1258
+ "dtype": "uint8"
1259
+ }
1260
+ ],
1261
+ "trajectory_sha256": "abdb2235f53d108e39588e8f8c3c53c408e060d3a3422c1f595a1c03f6682dc3",
1262
+ "source_sha256": "33cbc3331989dae93bd0bbe1b2c86d8631127a0de80945a604db8d567a893b9c",
1263
+ "view_sha256": "cce064142dab874ae0f47533ba53a03a81b7f46941bc6bf8532679736610a771"
1264
+ },
1265
+ {
1266
+ "demo_id": "demo_17",
1267
+ "file": "demo_17.h5",
1268
+ "length": 982,
1269
+ "fields": [
1270
+ {
1271
+ "path": "actions",
1272
+ "shape": [
1273
+ 982,
1274
+ 7
1275
+ ],
1276
+ "dtype": "float32"
1277
+ },
1278
+ {
1279
+ "path": "obs/agent/qpos",
1280
+ "shape": [
1281
+ 983,
1282
+ 9
1283
+ ],
1284
+ "dtype": "float32"
1285
+ },
1286
+ {
1287
+ "path": "obs/agent/qvel",
1288
+ "shape": [
1289
+ 983,
1290
+ 9
1291
+ ],
1292
+ "dtype": "float32"
1293
+ },
1294
+ {
1295
+ "path": "obs/sensor_data/front_camera/rgb",
1296
+ "shape": [
1297
+ 983,
1298
+ 240,
1299
+ 320,
1300
+ 3
1301
+ ],
1302
+ "dtype": "uint8"
1303
+ },
1304
+ {
1305
+ "path": "obs/sensor_data/left_camera/rgb",
1306
+ "shape": [
1307
+ 983,
1308
+ 240,
1309
+ 320,
1310
+ 3
1311
+ ],
1312
+ "dtype": "uint8"
1313
+ },
1314
+ {
1315
+ "path": "obs/sensor_data/robot_camera/rgb",
1316
+ "shape": [
1317
+ 983,
1318
+ 240,
1319
+ 320,
1320
+ 3
1321
+ ],
1322
+ "dtype": "uint8"
1323
+ },
1324
+ {
1325
+ "path": "obs/sensor_data/top_camera/rgb",
1326
+ "shape": [
1327
+ 983,
1328
+ 320,
1329
+ 240,
1330
+ 3
1331
+ ],
1332
+ "dtype": "uint8"
1333
+ }
1334
+ ],
1335
+ "trajectory_sha256": "c083cf71a9ed2df083b99caa5ef3ae2f2001bdda6d744210c7b8c7724ba24340",
1336
+ "source_sha256": "7497d94e4c251e76aaf634edfeb931fc590ee16e483c869ab16f46e00268e62e",
1337
+ "view_sha256": "5985e93304501fab390d847656863dd02420880fb56ad1701f904e62c852888a"
1338
+ },
1339
+ {
1340
+ "demo_id": "demo_18",
1341
+ "file": "demo_18.h5",
1342
+ "length": 963,
1343
+ "fields": [
1344
+ {
1345
+ "path": "actions",
1346
+ "shape": [
1347
+ 963,
1348
+ 7
1349
+ ],
1350
+ "dtype": "float32"
1351
+ },
1352
+ {
1353
+ "path": "obs/agent/qpos",
1354
+ "shape": [
1355
+ 964,
1356
+ 9
1357
+ ],
1358
+ "dtype": "float32"
1359
+ },
1360
+ {
1361
+ "path": "obs/agent/qvel",
1362
+ "shape": [
1363
+ 964,
1364
+ 9
1365
+ ],
1366
+ "dtype": "float32"
1367
+ },
1368
+ {
1369
+ "path": "obs/sensor_data/front_camera/rgb",
1370
+ "shape": [
1371
+ 964,
1372
+ 240,
1373
+ 320,
1374
+ 3
1375
+ ],
1376
+ "dtype": "uint8"
1377
+ },
1378
+ {
1379
+ "path": "obs/sensor_data/left_camera/rgb",
1380
+ "shape": [
1381
+ 964,
1382
+ 240,
1383
+ 320,
1384
+ 3
1385
+ ],
1386
+ "dtype": "uint8"
1387
+ },
1388
+ {
1389
+ "path": "obs/sensor_data/robot_camera/rgb",
1390
+ "shape": [
1391
+ 964,
1392
+ 240,
1393
+ 320,
1394
+ 3
1395
+ ],
1396
+ "dtype": "uint8"
1397
+ },
1398
+ {
1399
+ "path": "obs/sensor_data/top_camera/rgb",
1400
+ "shape": [
1401
+ 964,
1402
+ 320,
1403
+ 240,
1404
+ 3
1405
+ ],
1406
+ "dtype": "uint8"
1407
+ }
1408
+ ],
1409
+ "trajectory_sha256": "2051d2b363e9b765d39f55305f82b1322bdda6489abfdde8646cf65b3ba7c686",
1410
+ "source_sha256": "f28cb376cc628ea01a49ad906dacc21b2333e7016428dc9223b0a9a3ba625681",
1411
+ "view_sha256": "8eda9d9feefca942095002f341a46d0f069237d4579915d12e02e87d0eccaa5f"
1412
+ },
1413
+ {
1414
+ "demo_id": "demo_19",
1415
+ "file": "demo_19.h5",
1416
+ "length": 1000,
1417
+ "fields": [
1418
+ {
1419
+ "path": "actions",
1420
+ "shape": [
1421
+ 1000,
1422
+ 7
1423
+ ],
1424
+ "dtype": "float32"
1425
+ },
1426
+ {
1427
+ "path": "obs/agent/qpos",
1428
+ "shape": [
1429
+ 1001,
1430
+ 9
1431
+ ],
1432
+ "dtype": "float32"
1433
+ },
1434
+ {
1435
+ "path": "obs/agent/qvel",
1436
+ "shape": [
1437
+ 1001,
1438
+ 9
1439
+ ],
1440
+ "dtype": "float32"
1441
+ },
1442
+ {
1443
+ "path": "obs/sensor_data/front_camera/rgb",
1444
+ "shape": [
1445
+ 1001,
1446
+ 240,
1447
+ 320,
1448
+ 3
1449
+ ],
1450
+ "dtype": "uint8"
1451
+ },
1452
+ {
1453
+ "path": "obs/sensor_data/left_camera/rgb",
1454
+ "shape": [
1455
+ 1001,
1456
+ 240,
1457
+ 320,
1458
+ 3
1459
+ ],
1460
+ "dtype": "uint8"
1461
+ },
1462
+ {
1463
+ "path": "obs/sensor_data/robot_camera/rgb",
1464
+ "shape": [
1465
+ 1001,
1466
+ 240,
1467
+ 320,
1468
+ 3
1469
+ ],
1470
+ "dtype": "uint8"
1471
+ },
1472
+ {
1473
+ "path": "obs/sensor_data/top_camera/rgb",
1474
+ "shape": [
1475
+ 1001,
1476
+ 320,
1477
+ 240,
1478
+ 3
1479
+ ],
1480
+ "dtype": "uint8"
1481
+ }
1482
+ ],
1483
+ "trajectory_sha256": "0cb2ab11cc8401b2b077fc9bdc6a452ea557783c144a0194f0582e6954eabfa9",
1484
+ "source_sha256": "f24ebd7dc0b4d901463248c3037a1db989ac8e9bf1779f21a5f57ff4519a6f31",
1485
+ "view_sha256": "37bfd0098f066f42af58e55fb003ecf1b02c23454cca642b0d71e43d5da413e9"
1486
+ }
1487
+ ]
1488
+ }
agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/findings.md ADDED
@@ -0,0 +1,94 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Final candidate findings before independent rehearsals
2
+
3
+ Behavior derived from full interactive episode 16. Final guards reserve remaining reference actions before pausing, clear stale target confirmation when a new clamp lacks reference landmarks, and prevent pose mismatch from accelerating the reference clock or selecting future action frames. Final independent rehearsals 17–20 and runner freeze receipts are authoritative for this exact bundle; they are not yet claimed here. Baseline rehearsal 4 was valid with native success false.
4
+
5
+ Full interactive outcomes:
6
+ ```json
7
+ [
8
+ {
9
+ "sampled_top_frames": 39,
10
+ "sampled_top_dark_frames": 16,
11
+ "release_pause_calls": 31,
12
+ "live_grid_calls": 822,
13
+ "episode": 14,
14
+ "steps": 969,
15
+ "native_success_final": false,
16
+ "native_success_any_saved": false,
17
+ "peak_joint_error": 0.4269493818283081,
18
+ "visual_correction_calls": 373,
19
+ "selected_demo": "demo_12",
20
+ "initial_mask_error": 0.01751301996409893,
21
+ "final_image": "/workspace/exploration/ep14/final_top_camera.jpg",
22
+ "actions_saved": 968
23
+ },
24
+ {
25
+ "sampled_top_frames": 38,
26
+ "sampled_top_dark_frames": 7,
27
+ "release_pause_calls": 14,
28
+ "live_grid_calls": 877,
29
+ "episode": 15,
30
+ "steps": 942,
31
+ "native_success_final": false,
32
+ "native_success_any_saved": false,
33
+ "peak_joint_error": 0.6760162711143494,
34
+ "visual_correction_calls": 603,
35
+ "selected_demo": "demo_16",
36
+ "initial_mask_error": 0.011601562611758709,
37
+ "final_image": "/workspace/exploration/ep15/final_top_camera.jpg",
38
+ "actions_saved": 941
39
+ },
40
+ {
41
+ "sampled_top_frames": 22,
42
+ "sampled_top_dark_frames": 0,
43
+ "release_pause_calls": 0,
44
+ "live_grid_calls": 503,
45
+ "episode": 16,
46
+ "steps": 532,
47
+ "native_success_final": false,
48
+ "native_success_any_saved": false,
49
+ "peak_joint_error": 1.0243891477584839,
50
+ "visual_correction_calls": 483,
51
+ "selected_demo": "demo_08",
52
+ "initial_mask_error": 0.020494790747761726,
53
+ "final_image": "/workspace/exploration/ep16/final_top_camera.jpg",
54
+ "actions_saved": 531
55
+ }
56
+ ]
57
+ ```
58
+
59
+ The program has no verified semantic grasp or completion detector. Occlusion, incorrect landmarks, inferred post geometry, and lost cloth contact remain material failure modes. Stopping or wrist alignment is not task success.
60
+
61
+ # Live exploration ledger
62
+ Only this task's API and 20 authorized demonstrations are used. Budget first checked: 0/20. Native horizon 1000; initial API hold is step 1. Public visible rule: all four eyelets must unambiguously be threaded over posts with cloth pegged flat; uncertainty is false. Native sparse success, when supplied, is recorded without reinterpretation.
63
+
64
+ 1. Initial observation retrieval, 150 native steps. Program harness misread execute_actions wrapper after first action; latest_observation recovered step 2, no action replay. Closed grip with near-zero translations trapped retrieval in a stationary contact phase. Final top/front images show rotated cloth and unthreaded holes. Native success false. Initial pose and actions saved, except first issued action (deterministically reconstructible from initial candidate and start.pkl). Localization uses coarse appearance, no eyelet identity; grasp not established; no reliable recovery.
65
+ 2. Moving-action filter after 12 stationary retrievals, 160 steps. Robot progressed past contact plateau and cloth turned nearer axis-aligned. Top evidence still shows holes away from posts. Native success false. q and retrieval indices recorded. Static recovery changes motion but does not establish a correct grasp or post alignment.
66
+ 3. Testing initial cloth-silhouette choice plus local pose/frame alignment and bounded finite-difference joint-error correction. Uses only authorized numeric sequences and top frame 25; provenance in candidate/sequences_provenance.json. This is nonparametric, no neural training.
67
+
68
+ 3 result: full selected demo 06 sequence ended at step 388. Native success false. Final top image clearly shows several unthreaded eyelets. Median later joint tracking error 0.01486, maximum 0.05337; small pose error did not establish a correct grasp.
69
+ 4: independent exact baseline rehearsal, valid and finished; native_step_feedback success false. Bundle hash b61386bf43f45a335d4ac2e59ff2be7a90fe48ada417d439a619be8dc6c33360. API returned status/hash/native score, not camera images. No visible score substituted.
70
+ 5–7 preregistered comparison: fixed contour eyelet detection and Hough wrist tracking; clamp/release timing unchanged. Lateral correction bounds 0.02, 0.05, 0.08 normalized (2, 5, 8 mm requested per step). Trial horizon 190 to examine first clamp/transport/release, not to claim full task completion. Different fresh episodes confound direct gain comparison, so evidence remains qualitative. Every supplied observation retained as compressed numeric RGB/proprioception from episode 5 onward.
71
+
72
+ 5 result: demo 14 selected; 83/165 post-settling calls used visual correction, maximum joint error 0.17596. Final native success false; visible manipulated corner occluded, other holes off posts. Four-hole detector independently also ranks demo 14 closest.
73
+ 6 result: demo 16 selected; joint error 0.4303 at step 100. Reference wrist localization sometimes jumped from true wrist near (175,118) to an arm joint near (78,142), confirmed against authorized demo 16 frame 100. Final native success false. Correction gain alone cannot repair wrong landmarks.
74
+ Next corrective design: remove reference wrist samples >12 pixels from local median, interpolate gaps no longer than 15 frames, use median saturation/intensity in a 7x7 center patch, reduce pixel-error gain from /15 to /75, reject residuals above 20 pixels. These fixed rules do not modify neural weights.
75
+ 7 result: initial mask error 0.00370 (demo 03), but joint error reached 1.6668 during the clamp phase; native success false and final cloth remains rotated/off posts. Strong correction was not robust even with a close initial silhouette. A plausible contributor is the ill-conditioned local inverse mapping introducing rotational changes in response to translation error; next tests preserve demo rotation and limit inferred translation correction to 0.005.
76
+ 8–10 configuration: clean geometry, /75 pixel feedback, residual gate 20 pixels, lateral limits 0.015/0.025/0.04, translation-only joint correction bound 0.005, diagnostic horizon 300. Fixed changes motivated by observed failure modes; no success claim.
77
+ 8 result: native success false at 300; visual 197/275 post-settling calls; max joint error 0.27578. Large approach offset (~[-15,-17] pixels) caused the residual gate to suppress correction until clamp decay, a program logic failure. Follow-up gates raw wrist/reference consistency rather than the requested target offset, and preserves offset for the first 12 closed-gripper frames before decaying slowly.
78
+ 9 result: native success false at 300. Corrected motion remained less divergent than high-gain episode 7, but final images do not establish threaded eyelets. Native successes remain absent.
79
+ Post localization investigation: initial white-cap connected-component detector often produces no landmarks because caps connect to cloth. In episode 9 frames 140–175, one stable visible post is detected at approximately (187.56,134.30), but corresponding reference demo16 frames expose a different post; direct same-frame matching is invalid. Authorized demo16 top frame150 was additionally viewed (exploration/demo16_150.jpg). New offline experiment collects persistent cap detections through each authorized demo and requires a four-point square fit; it will be rejected if insufficiently constrained.
80
+ Reference post recovery: every demo has two dominant persistent cap clusters separated by 61–75 pixels; most are about 68. The two-point square inference chooses the side toward the initial cloth-hole centroid. Independent third/fourth observed caps in demos 02,07,11,13 agree within roughly 0.4–1.2 pixels. Additional distant weak clusters are treated as likely false cap detections, not ground truth. Explicit observed/inferred provenance is in posts_inferred_provenance.json. Episode 9's visible top-right post differs from demo16's inferred counterpart by about [18.3,13.4] pixels; matching only cloth can therefore choose the wrong transport direction.
81
+ Next trial program tracks cap detections only after spatially consistent repetition, with unambiguous nearest reference association (<28 pixels and 10-pixel margin). A confirmed target-post shift replaces the zero transport offset. No live post detection means the existing fallback remains; inferred reference geometry is never reported as observed task success. Grip-transition frame skipping is disabled within four frames of a switch.
82
+ 10 result: native success false at 300; arm stayed comparatively stable but cloth rotated away from target positions.\n11 result: native success false at 350, demo15 selected. No cap reached confirmation through most of the attempt, so the post-aware branch did not actually alter transport. This is a localization-coverage failure, not evidence against/for post correction itself.\n12 interim: a stable live post was confirmed during first clamp, producing target shift approximately [+24.72,-9.02] pixels relative to demo16. This confirms the transport-goal branch activates when legal cap evidence exists.\nUpcoming release version: circle-cap candidates include partially merged post tops, but require stationary repetition and geometric pair consistency before inferring a full live grid. Connected-component caps retain a direct-confirmation path. A native-call counter (independent of demo phase) enforces the horizon when pauses are added. At release, a confirmed target and a visible residual can hold the gripper closed for at most 30 extra steps while applying bounded XY corrections; it does not declare success.
83
+ Correction to episode12 interim interpretation: subsequent image inspection places the accepted [+24.7,-9.0] landmark near the gripper; it is not semantically confirmed as a post and may be a false robot feature. The phrase "confirmed" above referred only to repeated detector acceptance. Final release-version design rejects cap candidates within 18 pixels of the wrist and disables single-cap target shifts entirely: a stable two-cap grid fit is required. Ambiguous evidence is not counted as task success.
84
+ 12 result: native success false at 350; maximum joint error approximately 0.5595. Accepted single-cap landmark remains untrusted after image review.\n13 result: native success false at 350, demo06 selected. Stable cap counts for two left-side locations appeared and a second-grasp shift near [+7.62,+3.85] pixels was applied. Final image still shows several brown/unthreaded holes. Local phase alignment advanced to reference frame369 by native step350; grip-switch protection prevented skipping directly across switches, but frame advancement still changes phase duration.
85
+ 14–16: full-program interactive trials, up to the native 1000-step horizon, with final-style circle+component post tracking, wrist exclusion, two-cap grid fit, independent native-step counter, and bounded release pause. Lateral bounds 0.025/0.04/0.06. All observations/actions and the actual candidate files are retained per episode. These are still exploration, not independent rehearsal results.
86
+ 14 interim: live post square fitted by ~step150; first release pause active at native step200/reference178. After roughly step250, multiple supplied RGB streams became extremely dark or occluded (for example top frame300 has values 16–19, front frame300 has 1–20). Cause is not established from legal images; no simulator/state inspection was performed. Wrist localization falls back. These frames are failure/ambiguous evidence, not success. The full trial continues within the existing slot/horizon.
87
+ Final timing guards planned before independent rehearsals: extra pauses must leave enough native steps for every remaining reference action; stale target confirmation is cleared in closed phases without valid reference release landmarks. These are code guards, not new data or neural training.
88
+ 14 result: program finished at native step969, native success false throughout saved returns. 31 total release-pause calls across releases; max joint error 0.42695; 16 of 39 sampled top images have maximum intensity <40, with additional large occluding pale surfaces. No usable four-eyelet completion evidence.
89
+ 15 result: program finished at native step942, native success false throughout saved returns. Clear views lasted through about step525, then severe darkness/occlusion appeared after a later transport. The live square was stable, but wrist/post alignment did not establish successful cloth attachment. Reference-frame progress and native-step progress differ because of local alignment and release pauses.
90
+ Timing terminology: the 12-frame clamp rule retains the approach offset; visual corrections may still run. It is not a guarantee of a motionless gripper. No grasp flag or semantic threading detector is used.
91
+ 16 interim timing finding: reference frame487 was selected by native step450, a lead of ~37 frames with delta control. Local pose mismatch can therefore accelerate the reference clock and skip intended delta-motion duration. Hand/cloth images do not prove causation, but premature phase advance is a concrete program issue. Final revision will increment reference phase exactly once per normal action, permit at most a one-frame backward action alignment, and keep bounded release pauses. This change will be tested only through the required final independent rehearsals; no success is assumed.
92
+ 16 result: finished at native step532 while reaching the end of demo08's 573-action reference, native success false throughout saved returns. 0/22 sampled top frames were dark; visible-correction coverage was 483 calls; maximum joint error 1.02439. Final image still has clearly unthreaded holes and deformation. No successful grasp/threading capability is claimed.
93
+
94
+ Final selection before slots17–20: use episode16's behavior as the base because it preserved usable visual evidence throughout its full attempt, unlike episodes14–15. All three had zero native successes, and fresh-scene differences confound gain comparisons; this is a qualitative stability choice, not a demonstrated success-rate improvement. Apply the documented reference-clock, horizon-reservation, and stale-target guards. These guard changes have not received an extra interactive rollout; the four remaining authorized slots are their exact independent rehearsals. No additional rollout is permitted after slot20.
agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/geometry_clean_provenance.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "source": "fixed top-camera features from the same 20 authorized demonstrations; geometry_provenance.json",
3
+ "cleaning": "reject >12-pixel deviation from +/-7 frame local median; interpolate internal gaps <=15 frames; no model training"
4
+ }
agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/geometry_provenance.json ADDED
@@ -0,0 +1,182 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "demo": "demo_00",
4
+ "camera": "top_camera",
5
+ "frames": [
6
+ 0,
7
+ 558
8
+ ],
9
+ "processing": "fixed contour and Hough geometry"
10
+ },
11
+ {
12
+ "demo": "demo_01",
13
+ "camera": "top_camera",
14
+ "frames": [
15
+ 0,
16
+ 955
17
+ ],
18
+ "processing": "fixed contour and Hough geometry"
19
+ },
20
+ {
21
+ "demo": "demo_02",
22
+ "camera": "top_camera",
23
+ "frames": [
24
+ 0,
25
+ 615
26
+ ],
27
+ "processing": "fixed contour and Hough geometry"
28
+ },
29
+ {
30
+ "demo": "demo_03",
31
+ "camera": "top_camera",
32
+ "frames": [
33
+ 0,
34
+ 399
35
+ ],
36
+ "processing": "fixed contour and Hough geometry"
37
+ },
38
+ {
39
+ "demo": "demo_04",
40
+ "camera": "top_camera",
41
+ "frames": [
42
+ 0,
43
+ 652
44
+ ],
45
+ "processing": "fixed contour and Hough geometry"
46
+ },
47
+ {
48
+ "demo": "demo_05",
49
+ "camera": "top_camera",
50
+ "frames": [
51
+ 0,
52
+ 474
53
+ ],
54
+ "processing": "fixed contour and Hough geometry"
55
+ },
56
+ {
57
+ "demo": "demo_06",
58
+ "camera": "top_camera",
59
+ "frames": [
60
+ 0,
61
+ 387
62
+ ],
63
+ "processing": "fixed contour and Hough geometry"
64
+ },
65
+ {
66
+ "demo": "demo_07",
67
+ "camera": "top_camera",
68
+ "frames": [
69
+ 0,
70
+ 999
71
+ ],
72
+ "processing": "fixed contour and Hough geometry"
73
+ },
74
+ {
75
+ "demo": "demo_08",
76
+ "camera": "top_camera",
77
+ "frames": [
78
+ 0,
79
+ 572
80
+ ],
81
+ "processing": "fixed contour and Hough geometry"
82
+ },
83
+ {
84
+ "demo": "demo_09",
85
+ "camera": "top_camera",
86
+ "frames": [
87
+ 0,
88
+ 717
89
+ ],
90
+ "processing": "fixed contour and Hough geometry"
91
+ },
92
+ {
93
+ "demo": "demo_10",
94
+ "camera": "top_camera",
95
+ "frames": [
96
+ 0,
97
+ 473
98
+ ],
99
+ "processing": "fixed contour and Hough geometry"
100
+ },
101
+ {
102
+ "demo": "demo_11",
103
+ "camera": "top_camera",
104
+ "frames": [
105
+ 0,
106
+ 902
107
+ ],
108
+ "processing": "fixed contour and Hough geometry"
109
+ },
110
+ {
111
+ "demo": "demo_12",
112
+ "camera": "top_camera",
113
+ "frames": [
114
+ 0,
115
+ 941
116
+ ],
117
+ "processing": "fixed contour and Hough geometry"
118
+ },
119
+ {
120
+ "demo": "demo_13",
121
+ "camera": "top_camera",
122
+ "frames": [
123
+ 0,
124
+ 880
125
+ ],
126
+ "processing": "fixed contour and Hough geometry"
127
+ },
128
+ {
129
+ "demo": "demo_14",
130
+ "camera": "top_camera",
131
+ "frames": [
132
+ 0,
133
+ 602
134
+ ],
135
+ "processing": "fixed contour and Hough geometry"
136
+ },
137
+ {
138
+ "demo": "demo_15",
139
+ "camera": "top_camera",
140
+ "frames": [
141
+ 0,
142
+ 722
143
+ ],
144
+ "processing": "fixed contour and Hough geometry"
145
+ },
146
+ {
147
+ "demo": "demo_16",
148
+ "camera": "top_camera",
149
+ "frames": [
150
+ 0,
151
+ 953
152
+ ],
153
+ "processing": "fixed contour and Hough geometry"
154
+ },
155
+ {
156
+ "demo": "demo_17",
157
+ "camera": "top_camera",
158
+ "frames": [
159
+ 0,
160
+ 981
161
+ ],
162
+ "processing": "fixed contour and Hough geometry"
163
+ },
164
+ {
165
+ "demo": "demo_18",
166
+ "camera": "top_camera",
167
+ "frames": [
168
+ 0,
169
+ 962
170
+ ],
171
+ "processing": "fixed contour and Hough geometry"
172
+ },
173
+ {
174
+ "demo": "demo_19",
175
+ "camera": "top_camera",
176
+ "frames": [
177
+ 0,
178
+ 999
179
+ ],
180
+ "processing": "fixed contour and Hough geometry"
181
+ }
182
+ ]
agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/handoff.md ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ # Offline final handoff
2
+ This immutable candidate is prepared for independent rehearsals 17–20. The exact rehearsed bundle hash and selected artifact are supplied by the runner; consult exploration/rehearsal_*.json and exploration/freeze.json outside this bundle for final receipts.
3
+
4
+ Entry points: controller.py reset_episode() and act(obs). Runtime: Python, NumPy, OpenCV, sequences.npz, geometry_clean.npz, posts_inferred.npz. All code and nonparametric assets are local and offline. No network, model-development agent, external model, simulator, worker API, or demonstration mount is required by the controller. No neural weights were trained or loaded.
5
+
6
+ Input is the legal observation dictionary with top_camera RGB and agent/qpos under proprioception; step_id is used when supplied, otherwise an internal native-call counter includes the runner hold. Output is one bounded Panda action of width seven, or None at reference-program end/native horizon. A stop is not a success assertion. Additional release pauses count against the same native horizon.
7
+
8
+ All 20 authorized demonstrations were analyzed; no other-task material was used. Reference post squares are explicitly inferred from persistent legal-image caps, not simulator state. Runtime live-square inference requires consistent repeated detections. No reliable grasp or four-eyelet completion detector exists. Failures and all exploration snapshots remain in the task workspace. Native sparse feedback takes precedence; absent native success, ambiguous visual evidence counts false.
9
+
10
+ The runner supplies the verified demonstration manifest, read-only authorized demonstrations, and common environment lock. All final rehearsal slots must be consumed before freeze. Do not modify behavior after the last rehearsal and do not run a 21st episode.
agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/policy_runner.py ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Frozen-program entry point; no development agent and no reset capability."""
2
+ import importlib.util
3
+ import sys
4
+
5
+ sys.path.insert(0,'/artifact')
6
+ sys.path.insert(0,'/')
7
+ from worker_api import API
8
+
9
+
10
+ def main():
11
+ spec=importlib.util.spec_from_file_location('controller','/artifact/controller.py')
12
+ module=importlib.util.module_from_spec(spec)
13
+ spec.loader.exec_module(module)
14
+ module.reset_episode()
15
+ api=API()
16
+ result=api.latest_observation()
17
+ while not (result['terminated'] or result['truncated'] or result['horizon_reached']):
18
+ obs=dict(result['observation'],step_id=result['step_id'])
19
+ action=module.act(obs)
20
+ if action is None:break
21
+ response=api.execute_actions([action])
22
+ result=response['steps'][-1]
23
+
24
+
25
+ if __name__=='__main__':main()
agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/posts_inferred_provenance.json ADDED
@@ -0,0 +1,3924 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "demo": "demo_00",
4
+ "processed_frames": [
5
+ 25,
6
+ 30,
7
+ 35,
8
+ 40,
9
+ 45,
10
+ 50,
11
+ 55,
12
+ 60,
13
+ 65,
14
+ 70,
15
+ 75,
16
+ 80,
17
+ 85,
18
+ 90,
19
+ 95,
20
+ 100,
21
+ 105,
22
+ 110,
23
+ 115,
24
+ 120,
25
+ 125,
26
+ 130,
27
+ 135,
28
+ 140,
29
+ 145,
30
+ 150,
31
+ 155,
32
+ 160,
33
+ 165,
34
+ 170,
35
+ 175,
36
+ 180,
37
+ 185,
38
+ 190,
39
+ 195,
40
+ 200,
41
+ 205,
42
+ 210,
43
+ 215,
44
+ 220,
45
+ 225,
46
+ 230,
47
+ 235,
48
+ 240,
49
+ 245,
50
+ 250,
51
+ 255,
52
+ 260,
53
+ 265,
54
+ 270,
55
+ 275,
56
+ 280,
57
+ 285,
58
+ 290,
59
+ 295,
60
+ 300,
61
+ 305,
62
+ 310,
63
+ 315,
64
+ 320,
65
+ 325,
66
+ 330,
67
+ 335,
68
+ 340,
69
+ 345,
70
+ 350,
71
+ 355,
72
+ 360,
73
+ 365,
74
+ 370,
75
+ 375,
76
+ 380,
77
+ 385,
78
+ 390,
79
+ 395,
80
+ 400,
81
+ 405,
82
+ 410,
83
+ 415,
84
+ 420,
85
+ 425,
86
+ 430,
87
+ 435,
88
+ 440,
89
+ 445,
90
+ 450,
91
+ 455,
92
+ 460,
93
+ 465,
94
+ 470,
95
+ 475,
96
+ 480,
97
+ 485,
98
+ 490,
99
+ 495,
100
+ 500,
101
+ 505,
102
+ 510,
103
+ 515,
104
+ 520,
105
+ 525,
106
+ 530,
107
+ 535,
108
+ 540,
109
+ 545,
110
+ 550,
111
+ 555
112
+ ],
113
+ "decoded_range": [
114
+ 0,
115
+ 558
116
+ ],
117
+ "clusters": [
118
+ [
119
+ [
120
+ 117.56341491104305,
121
+ 122.28822286013941
122
+ ],
123
+ 62
124
+ ],
125
+ [
126
+ [
127
+ 101.67784992784992,
128
+ 188.75322455322453
129
+ ],
130
+ 14
131
+ ]
132
+ ],
133
+ "selected": [],
134
+ "inferred_grid": [
135
+ [
136
+ 117.56341491104305,
137
+ 122.28822286013941
138
+ ],
139
+ [
140
+ 101.67784992784992,
141
+ 188.75322455322453
142
+ ],
143
+ [
144
+ 168.14285162093503,
145
+ 204.63878953641768
146
+ ],
147
+ [
148
+ 184.02841660412815,
149
+ 138.17378784333255
150
+ ]
151
+ ],
152
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
153
+ "observed_side_length": 68.33701504234041,
154
+ "other_cluster_residuals": []
155
+ },
156
+ {
157
+ "demo": "demo_01",
158
+ "processed_frames": [
159
+ 25,
160
+ 30,
161
+ 35,
162
+ 40,
163
+ 45,
164
+ 50,
165
+ 55,
166
+ 60,
167
+ 65,
168
+ 70,
169
+ 75,
170
+ 80,
171
+ 85,
172
+ 90,
173
+ 95,
174
+ 100,
175
+ 105,
176
+ 110,
177
+ 115,
178
+ 120,
179
+ 125,
180
+ 130,
181
+ 135,
182
+ 140,
183
+ 145,
184
+ 150,
185
+ 155,
186
+ 160,
187
+ 165,
188
+ 170,
189
+ 175,
190
+ 180,
191
+ 185,
192
+ 190,
193
+ 195,
194
+ 200,
195
+ 205,
196
+ 210,
197
+ 215,
198
+ 220,
199
+ 225,
200
+ 230,
201
+ 235,
202
+ 240,
203
+ 245,
204
+ 250,
205
+ 255,
206
+ 260,
207
+ 265,
208
+ 270,
209
+ 275,
210
+ 280,
211
+ 285,
212
+ 290,
213
+ 295,
214
+ 300,
215
+ 305,
216
+ 310,
217
+ 315,
218
+ 320,
219
+ 325,
220
+ 330,
221
+ 335,
222
+ 340,
223
+ 345,
224
+ 350,
225
+ 355,
226
+ 360,
227
+ 365,
228
+ 370,
229
+ 375,
230
+ 380,
231
+ 385,
232
+ 390,
233
+ 395,
234
+ 400,
235
+ 405,
236
+ 410,
237
+ 415,
238
+ 420,
239
+ 425,
240
+ 430,
241
+ 435,
242
+ 440,
243
+ 445,
244
+ 450,
245
+ 455,
246
+ 460,
247
+ 465,
248
+ 470,
249
+ 475,
250
+ 480,
251
+ 485,
252
+ 490,
253
+ 495,
254
+ 500,
255
+ 505,
256
+ 510,
257
+ 515,
258
+ 520,
259
+ 525,
260
+ 530,
261
+ 535,
262
+ 540,
263
+ 545,
264
+ 550,
265
+ 555,
266
+ 560,
267
+ 565,
268
+ 570,
269
+ 575,
270
+ 580,
271
+ 585,
272
+ 590,
273
+ 595,
274
+ 600,
275
+ 605,
276
+ 610,
277
+ 615,
278
+ 620,
279
+ 625,
280
+ 630,
281
+ 635,
282
+ 640,
283
+ 645,
284
+ 650,
285
+ 655,
286
+ 660,
287
+ 665,
288
+ 670,
289
+ 675,
290
+ 680,
291
+ 685,
292
+ 690,
293
+ 695,
294
+ 700,
295
+ 705,
296
+ 710,
297
+ 715,
298
+ 720,
299
+ 725,
300
+ 730,
301
+ 735,
302
+ 740,
303
+ 745,
304
+ 750,
305
+ 755,
306
+ 760,
307
+ 765,
308
+ 770,
309
+ 775,
310
+ 780,
311
+ 785,
312
+ 790,
313
+ 795,
314
+ 800,
315
+ 805,
316
+ 810,
317
+ 815,
318
+ 820,
319
+ 825,
320
+ 830,
321
+ 835,
322
+ 840,
323
+ 845,
324
+ 850,
325
+ 855,
326
+ 860,
327
+ 865,
328
+ 870,
329
+ 875,
330
+ 880,
331
+ 885,
332
+ 890,
333
+ 895,
334
+ 900,
335
+ 905,
336
+ 910,
337
+ 915,
338
+ 920,
339
+ 925,
340
+ 930,
341
+ 935,
342
+ 940,
343
+ 945,
344
+ 950,
345
+ 955
346
+ ],
347
+ "decoded_range": [
348
+ 0,
349
+ 955
350
+ ],
351
+ "clusters": [
352
+ [
353
+ [
354
+ 170.56871021988886,
355
+ 118.1527440892846
356
+ ],
357
+ 72
358
+ ],
359
+ [
360
+ [
361
+ 103.43923750323752,
362
+ 130.56231649831642
363
+ ],
364
+ 25
365
+ ],
366
+ [
367
+ [
368
+ 166.39720524243018,
369
+ 110.65277445318338
370
+ ],
371
+ 15
372
+ ]
373
+ ],
374
+ "selected": [],
375
+ "inferred_grid": [
376
+ [
377
+ 170.56871021988886,
378
+ 118.1527440892846
379
+ ],
380
+ [
381
+ 103.43923750323752,
382
+ 130.56231649831642
383
+ ],
384
+ [
385
+ 115.84880991226935,
386
+ 197.69178921496774
387
+ ],
388
+ [
389
+ 182.97828262892068,
390
+ 185.28221680593595
391
+ ]
392
+ ],
393
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
394
+ "observed_side_length": 68.26685575438964,
395
+ "other_cluster_residuals": [
396
+ 8.582015982238829
397
+ ]
398
+ },
399
+ {
400
+ "demo": "demo_02",
401
+ "processed_frames": [
402
+ 25,
403
+ 30,
404
+ 35,
405
+ 40,
406
+ 45,
407
+ 50,
408
+ 55,
409
+ 60,
410
+ 65,
411
+ 70,
412
+ 75,
413
+ 80,
414
+ 85,
415
+ 90,
416
+ 95,
417
+ 100,
418
+ 105,
419
+ 110,
420
+ 115,
421
+ 120,
422
+ 125,
423
+ 130,
424
+ 135,
425
+ 140,
426
+ 145,
427
+ 150,
428
+ 155,
429
+ 160,
430
+ 165,
431
+ 170,
432
+ 175,
433
+ 180,
434
+ 185,
435
+ 190,
436
+ 195,
437
+ 200,
438
+ 205,
439
+ 210,
440
+ 215,
441
+ 220,
442
+ 225,
443
+ 230,
444
+ 235,
445
+ 240,
446
+ 245,
447
+ 250,
448
+ 255,
449
+ 260,
450
+ 265,
451
+ 270,
452
+ 275,
453
+ 280,
454
+ 285,
455
+ 290,
456
+ 295,
457
+ 300,
458
+ 305,
459
+ 310,
460
+ 315,
461
+ 320,
462
+ 325,
463
+ 330,
464
+ 335,
465
+ 340,
466
+ 345,
467
+ 350,
468
+ 355,
469
+ 360,
470
+ 365,
471
+ 370,
472
+ 375,
473
+ 380,
474
+ 385,
475
+ 390,
476
+ 395,
477
+ 400,
478
+ 405,
479
+ 410,
480
+ 415,
481
+ 420,
482
+ 425,
483
+ 430,
484
+ 435,
485
+ 440,
486
+ 445,
487
+ 450,
488
+ 455,
489
+ 460,
490
+ 465,
491
+ 470,
492
+ 475,
493
+ 480,
494
+ 485,
495
+ 490,
496
+ 495,
497
+ 500,
498
+ 505,
499
+ 510,
500
+ 515,
501
+ 520,
502
+ 525,
503
+ 530,
504
+ 535,
505
+ 540,
506
+ 545,
507
+ 550,
508
+ 555,
509
+ 560,
510
+ 565,
511
+ 570,
512
+ 575,
513
+ 580,
514
+ 585,
515
+ 590,
516
+ 595,
517
+ 600,
518
+ 605,
519
+ 610,
520
+ 615
521
+ ],
522
+ "decoded_range": [
523
+ 0,
524
+ 615
525
+ ],
526
+ "clusters": [
527
+ [
528
+ [
529
+ 111.45009923161183,
530
+ 195.00205334659114
531
+ ],
532
+ 49
533
+ ],
534
+ [
535
+ [
536
+ 179.24072398190035,
537
+ 186.62254901960785
538
+ ],
539
+ 26
540
+ ],
541
+ [
542
+ [
543
+ 102.68197772936338,
544
+ 127.29181796175259
545
+ ],
546
+ 18
547
+ ]
548
+ ],
549
+ "selected": [],
550
+ "inferred_grid": [
551
+ [
552
+ 111.45009923161183,
553
+ 195.00205334659114
554
+ ],
555
+ [
556
+ 179.24072398190035,
557
+ 186.62254901960785
558
+ ],
559
+ [
560
+ 170.86121965491705,
561
+ 118.83192426931933
562
+ ],
563
+ [
564
+ 103.07059490462854,
565
+ 127.21142859630262
566
+ ]
567
+ ],
568
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
569
+ "observed_side_length": 68.30655090692518,
570
+ "other_cluster_residuals": [
571
+ 0.39684475426609117
572
+ ]
573
+ },
574
+ {
575
+ "demo": "demo_03",
576
+ "processed_frames": [
577
+ 25,
578
+ 30,
579
+ 35,
580
+ 40,
581
+ 45,
582
+ 50,
583
+ 55,
584
+ 60,
585
+ 65,
586
+ 70,
587
+ 75,
588
+ 80,
589
+ 85,
590
+ 90,
591
+ 95,
592
+ 100,
593
+ 105,
594
+ 110,
595
+ 115,
596
+ 120,
597
+ 125,
598
+ 130,
599
+ 135,
600
+ 140,
601
+ 145,
602
+ 150,
603
+ 155,
604
+ 160,
605
+ 165,
606
+ 170,
607
+ 175,
608
+ 180,
609
+ 185,
610
+ 190,
611
+ 195,
612
+ 200,
613
+ 205,
614
+ 210,
615
+ 215,
616
+ 220,
617
+ 225,
618
+ 230,
619
+ 235,
620
+ 240,
621
+ 245,
622
+ 250,
623
+ 255,
624
+ 260,
625
+ 265,
626
+ 270,
627
+ 275,
628
+ 280,
629
+ 285,
630
+ 290,
631
+ 295,
632
+ 300,
633
+ 305,
634
+ 310,
635
+ 315,
636
+ 320,
637
+ 325,
638
+ 330,
639
+ 335,
640
+ 340,
641
+ 345,
642
+ 350,
643
+ 355,
644
+ 360,
645
+ 365,
646
+ 370,
647
+ 375,
648
+ 380,
649
+ 385,
650
+ 390,
651
+ 395
652
+ ],
653
+ "decoded_range": [
654
+ 0,
655
+ 399
656
+ ],
657
+ "clusters": [
658
+ [
659
+ [
660
+ 178.0444444444444,
661
+ 192.7111111111111
662
+ ],
663
+ 55
664
+ ],
665
+ [
666
+ [
667
+ 109.52822050315788,
668
+ 191.41160922990497
669
+ ],
670
+ 21
671
+ ]
672
+ ],
673
+ "selected": [],
674
+ "inferred_grid": [
675
+ [
676
+ 178.0444444444444,
677
+ 192.7111111111111
678
+ ],
679
+ [
680
+ 109.52822050315788,
681
+ 191.41160922990497
682
+ ],
683
+ [
684
+ 110.82772238436402,
685
+ 122.89538528861846
686
+ ],
687
+ [
688
+ 179.34394632565053,
689
+ 124.1948871698246
690
+ ]
691
+ ],
692
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
693
+ "observed_side_length": 68.52854622937642,
694
+ "other_cluster_residuals": []
695
+ },
696
+ {
697
+ "demo": "demo_04",
698
+ "processed_frames": [
699
+ 25,
700
+ 30,
701
+ 35,
702
+ 40,
703
+ 45,
704
+ 50,
705
+ 55,
706
+ 60,
707
+ 65,
708
+ 70,
709
+ 75,
710
+ 80,
711
+ 85,
712
+ 90,
713
+ 95,
714
+ 100,
715
+ 105,
716
+ 110,
717
+ 115,
718
+ 120,
719
+ 125,
720
+ 130,
721
+ 135,
722
+ 140,
723
+ 145,
724
+ 150,
725
+ 155,
726
+ 160,
727
+ 165,
728
+ 170,
729
+ 175,
730
+ 180,
731
+ 185,
732
+ 190,
733
+ 195,
734
+ 200,
735
+ 205,
736
+ 210,
737
+ 215,
738
+ 220,
739
+ 225,
740
+ 230,
741
+ 235,
742
+ 240,
743
+ 245,
744
+ 250,
745
+ 255,
746
+ 260,
747
+ 265,
748
+ 270,
749
+ 275,
750
+ 280,
751
+ 285,
752
+ 290,
753
+ 295,
754
+ 300,
755
+ 305,
756
+ 310,
757
+ 315,
758
+ 320,
759
+ 325,
760
+ 330,
761
+ 335,
762
+ 340,
763
+ 345,
764
+ 350,
765
+ 355,
766
+ 360,
767
+ 365,
768
+ 370,
769
+ 375,
770
+ 380,
771
+ 385,
772
+ 390,
773
+ 395,
774
+ 400,
775
+ 405,
776
+ 410,
777
+ 415,
778
+ 420,
779
+ 425,
780
+ 430,
781
+ 435,
782
+ 440,
783
+ 445,
784
+ 450,
785
+ 455,
786
+ 460,
787
+ 465,
788
+ 470,
789
+ 475,
790
+ 480,
791
+ 485,
792
+ 490,
793
+ 495,
794
+ 500,
795
+ 505,
796
+ 510,
797
+ 515,
798
+ 520,
799
+ 525,
800
+ 530,
801
+ 535,
802
+ 540,
803
+ 545,
804
+ 550,
805
+ 555,
806
+ 560,
807
+ 565,
808
+ 570,
809
+ 575,
810
+ 580,
811
+ 585,
812
+ 590,
813
+ 595,
814
+ 600,
815
+ 605,
816
+ 610,
817
+ 615,
818
+ 620,
819
+ 625,
820
+ 630,
821
+ 635,
822
+ 640,
823
+ 645,
824
+ 650
825
+ ],
826
+ "decoded_range": [
827
+ 0,
828
+ 652
829
+ ],
830
+ "clusters": [
831
+ [
832
+ [
833
+ 106.76844390177708,
834
+ 194.29879256545922
835
+ ],
836
+ 84
837
+ ],
838
+ [
839
+ [
840
+ 175.22222222222234,
841
+ 197.51851851851845
842
+ ],
843
+ 36
844
+ ]
845
+ ],
846
+ "selected": [],
847
+ "inferred_grid": [
848
+ [
849
+ 106.76844390177708,
850
+ 194.29879256545922
851
+ ],
852
+ [
853
+ 175.22222222222234,
854
+ 197.51851851851845
855
+ ],
856
+ [
857
+ 178.44194817528157,
858
+ 129.0647401980732
859
+ ],
860
+ [
861
+ 109.9881698548363,
862
+ 125.84501424501396
863
+ ]
864
+ ],
865
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
866
+ "observed_side_length": 68.52945645164176,
867
+ "other_cluster_residuals": []
868
+ },
869
+ {
870
+ "demo": "demo_05",
871
+ "processed_frames": [
872
+ 25,
873
+ 30,
874
+ 35,
875
+ 40,
876
+ 45,
877
+ 50,
878
+ 55,
879
+ 60,
880
+ 65,
881
+ 70,
882
+ 75,
883
+ 80,
884
+ 85,
885
+ 90,
886
+ 95,
887
+ 100,
888
+ 105,
889
+ 110,
890
+ 115,
891
+ 120,
892
+ 125,
893
+ 130,
894
+ 135,
895
+ 140,
896
+ 145,
897
+ 150,
898
+ 155,
899
+ 160,
900
+ 165,
901
+ 170,
902
+ 175,
903
+ 180,
904
+ 185,
905
+ 190,
906
+ 195,
907
+ 200,
908
+ 205,
909
+ 210,
910
+ 215,
911
+ 220,
912
+ 225,
913
+ 230,
914
+ 235,
915
+ 240,
916
+ 245,
917
+ 250,
918
+ 255,
919
+ 260,
920
+ 265,
921
+ 270,
922
+ 275,
923
+ 280,
924
+ 285,
925
+ 290,
926
+ 295,
927
+ 300,
928
+ 305,
929
+ 310,
930
+ 315,
931
+ 320,
932
+ 325,
933
+ 330,
934
+ 335,
935
+ 340,
936
+ 345,
937
+ 350,
938
+ 355,
939
+ 360,
940
+ 365,
941
+ 370,
942
+ 375,
943
+ 380,
944
+ 385,
945
+ 390,
946
+ 395,
947
+ 400,
948
+ 405,
949
+ 410,
950
+ 415,
951
+ 420,
952
+ 425,
953
+ 430,
954
+ 435,
955
+ 440,
956
+ 445,
957
+ 450,
958
+ 455,
959
+ 460,
960
+ 465,
961
+ 470
962
+ ],
963
+ "decoded_range": [
964
+ 0,
965
+ 474
966
+ ],
967
+ "clusters": [
968
+ [
969
+ [
970
+ 108.74509803921562,
971
+ 131.84313725490202
972
+ ],
973
+ 45
974
+ ],
975
+ [
976
+ [
977
+ 114.65137721755366,
978
+ 200.07280451316814
979
+ ],
980
+ 28
981
+ ]
982
+ ],
983
+ "selected": [],
984
+ "inferred_grid": [
985
+ [
986
+ 108.74509803921562,
987
+ 131.84313725490202
988
+ ],
989
+ [
990
+ 114.65137721755366,
991
+ 200.07280451316814
992
+ ],
993
+ [
994
+ 182.88104447581978,
995
+ 194.1665253348301
996
+ ],
997
+ [
998
+ 176.97476529748172,
999
+ 125.93685807656398
1000
+ ]
1001
+ ],
1002
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
1003
+ "observed_side_length": 68.4848277204972,
1004
+ "other_cluster_residuals": []
1005
+ },
1006
+ {
1007
+ "demo": "demo_06",
1008
+ "processed_frames": [
1009
+ 25,
1010
+ 30,
1011
+ 35,
1012
+ 40,
1013
+ 45,
1014
+ 50,
1015
+ 55,
1016
+ 60,
1017
+ 65,
1018
+ 70,
1019
+ 75,
1020
+ 80,
1021
+ 85,
1022
+ 90,
1023
+ 95,
1024
+ 100,
1025
+ 105,
1026
+ 110,
1027
+ 115,
1028
+ 120,
1029
+ 125,
1030
+ 130,
1031
+ 135,
1032
+ 140,
1033
+ 145,
1034
+ 150,
1035
+ 155,
1036
+ 160,
1037
+ 165,
1038
+ 170,
1039
+ 175,
1040
+ 180,
1041
+ 185,
1042
+ 190,
1043
+ 195,
1044
+ 200,
1045
+ 205,
1046
+ 210,
1047
+ 215,
1048
+ 220,
1049
+ 225,
1050
+ 230,
1051
+ 235,
1052
+ 240,
1053
+ 245,
1054
+ 250,
1055
+ 255,
1056
+ 260,
1057
+ 265,
1058
+ 270,
1059
+ 275,
1060
+ 280,
1061
+ 285,
1062
+ 290,
1063
+ 295,
1064
+ 300,
1065
+ 305,
1066
+ 310,
1067
+ 315,
1068
+ 320,
1069
+ 325,
1070
+ 330,
1071
+ 335,
1072
+ 340,
1073
+ 345,
1074
+ 350,
1075
+ 355,
1076
+ 360,
1077
+ 365,
1078
+ 370,
1079
+ 375,
1080
+ 380,
1081
+ 385
1082
+ ],
1083
+ "decoded_range": [
1084
+ 0,
1085
+ 387
1086
+ ],
1087
+ "clusters": [
1088
+ [
1089
+ [
1090
+ 103.55555555555564,
1091
+ 124.29629629629629
1092
+ ],
1093
+ 44
1094
+ ],
1095
+ [
1096
+ [
1097
+ 105.7826193130541,
1098
+ 192.68253968253967
1099
+ ],
1100
+ 23
1101
+ ]
1102
+ ],
1103
+ "selected": [],
1104
+ "inferred_grid": [
1105
+ [
1106
+ 103.55555555555564,
1107
+ 124.29629629629629
1108
+ ],
1109
+ [
1110
+ 105.7826193130541,
1111
+ 192.68253968253967
1112
+ ],
1113
+ [
1114
+ 174.16886269929748,
1115
+ 190.4554759250412
1116
+ ],
1117
+ [
1118
+ 171.94179894179902,
1119
+ 122.06923253879783
1120
+ ]
1121
+ ],
1122
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
1123
+ "observed_side_length": 68.42249701276971,
1124
+ "other_cluster_residuals": []
1125
+ },
1126
+ {
1127
+ "demo": "demo_07",
1128
+ "processed_frames": [
1129
+ 25,
1130
+ 30,
1131
+ 35,
1132
+ 40,
1133
+ 45,
1134
+ 50,
1135
+ 55,
1136
+ 60,
1137
+ 65,
1138
+ 70,
1139
+ 75,
1140
+ 80,
1141
+ 85,
1142
+ 90,
1143
+ 95,
1144
+ 100,
1145
+ 105,
1146
+ 110,
1147
+ 115,
1148
+ 120,
1149
+ 125,
1150
+ 130,
1151
+ 135,
1152
+ 140,
1153
+ 145,
1154
+ 150,
1155
+ 155,
1156
+ 160,
1157
+ 165,
1158
+ 170,
1159
+ 175,
1160
+ 180,
1161
+ 185,
1162
+ 190,
1163
+ 195,
1164
+ 200,
1165
+ 205,
1166
+ 210,
1167
+ 215,
1168
+ 220,
1169
+ 225,
1170
+ 230,
1171
+ 235,
1172
+ 240,
1173
+ 245,
1174
+ 250,
1175
+ 255,
1176
+ 260,
1177
+ 265,
1178
+ 270,
1179
+ 275,
1180
+ 280,
1181
+ 285,
1182
+ 290,
1183
+ 295,
1184
+ 300,
1185
+ 305,
1186
+ 310,
1187
+ 315,
1188
+ 320,
1189
+ 325,
1190
+ 330,
1191
+ 335,
1192
+ 340,
1193
+ 345,
1194
+ 350,
1195
+ 355,
1196
+ 360,
1197
+ 365,
1198
+ 370,
1199
+ 375,
1200
+ 380,
1201
+ 385,
1202
+ 390,
1203
+ 395,
1204
+ 400,
1205
+ 405,
1206
+ 410,
1207
+ 415,
1208
+ 420,
1209
+ 425,
1210
+ 430,
1211
+ 435,
1212
+ 440,
1213
+ 445,
1214
+ 450,
1215
+ 455,
1216
+ 460,
1217
+ 465,
1218
+ 470,
1219
+ 475,
1220
+ 480,
1221
+ 485,
1222
+ 490,
1223
+ 495,
1224
+ 500,
1225
+ 505,
1226
+ 510,
1227
+ 515,
1228
+ 520,
1229
+ 525,
1230
+ 530,
1231
+ 535,
1232
+ 540,
1233
+ 545,
1234
+ 550,
1235
+ 555,
1236
+ 560,
1237
+ 565,
1238
+ 570,
1239
+ 575,
1240
+ 580,
1241
+ 585,
1242
+ 590,
1243
+ 595,
1244
+ 600,
1245
+ 605,
1246
+ 610,
1247
+ 615,
1248
+ 620,
1249
+ 625,
1250
+ 630,
1251
+ 635,
1252
+ 640,
1253
+ 645,
1254
+ 650,
1255
+ 655,
1256
+ 660,
1257
+ 665,
1258
+ 670,
1259
+ 675,
1260
+ 680,
1261
+ 685,
1262
+ 690,
1263
+ 695,
1264
+ 700,
1265
+ 705,
1266
+ 710,
1267
+ 715,
1268
+ 720,
1269
+ 725,
1270
+ 730,
1271
+ 735,
1272
+ 740,
1273
+ 745,
1274
+ 750,
1275
+ 755,
1276
+ 760,
1277
+ 765,
1278
+ 770,
1279
+ 775,
1280
+ 780,
1281
+ 785,
1282
+ 790,
1283
+ 795,
1284
+ 800,
1285
+ 805,
1286
+ 810,
1287
+ 815,
1288
+ 820,
1289
+ 825,
1290
+ 830,
1291
+ 835,
1292
+ 840,
1293
+ 845,
1294
+ 850,
1295
+ 855,
1296
+ 860,
1297
+ 865,
1298
+ 870,
1299
+ 875,
1300
+ 880,
1301
+ 885,
1302
+ 890,
1303
+ 895,
1304
+ 900,
1305
+ 905,
1306
+ 910,
1307
+ 915,
1308
+ 920,
1309
+ 925,
1310
+ 930,
1311
+ 935,
1312
+ 940,
1313
+ 945,
1314
+ 950,
1315
+ 955,
1316
+ 960,
1317
+ 965,
1318
+ 970,
1319
+ 975,
1320
+ 980,
1321
+ 985,
1322
+ 990,
1323
+ 995
1324
+ ],
1325
+ "decoded_range": [
1326
+ 0,
1327
+ 999
1328
+ ],
1329
+ "clusters": [
1330
+ [
1331
+ [
1332
+ 180.9375,
1333
+ 190.77083333333317
1334
+ ],
1335
+ 120
1336
+ ],
1337
+ [
1338
+ [
1339
+ 112.40838206627681,
1340
+ 194.79727095516577
1341
+ ],
1342
+ 19
1343
+ ],
1344
+ [
1345
+ [
1346
+ 177.49895334928226,
1347
+ 122.57012559808615
1348
+ ],
1349
+ 16
1350
+ ],
1351
+ [
1352
+ [
1353
+ 109.37522281639929,
1354
+ 125.71657754010694
1355
+ ],
1356
+ 8
1357
+ ]
1358
+ ],
1359
+ "selected": [
1360
+ [
1361
+ 109.37522281639929,
1362
+ 125.71657754010694
1363
+ ],
1364
+ [
1365
+ 177.49895334928226,
1366
+ 122.57012559808615
1367
+ ],
1368
+ [
1369
+ 180.9375,
1370
+ 190.77083333333317
1371
+ ],
1372
+ [
1373
+ 112.40838206627681,
1374
+ 194.79727095516577
1375
+ ]
1376
+ ],
1377
+ "inferred_grid": [
1378
+ [
1379
+ 180.9375,
1380
+ 190.77083333333317
1381
+ ],
1382
+ [
1383
+ 112.40838206627681,
1384
+ 194.79727095516577
1385
+ ],
1386
+ [
1387
+ 108.38194444444422,
1388
+ 126.26815302144259
1389
+ ],
1390
+ [
1391
+ 176.9110623781674,
1392
+ 122.24171539960999
1393
+ ]
1394
+ ],
1395
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
1396
+ "observed_side_length": 68.64730296739013,
1397
+ "other_cluster_residuals": [
1398
+ 0.6734011080934791,
1399
+ 1.1361502699046315
1400
+ ]
1401
+ },
1402
+ {
1403
+ "demo": "demo_08",
1404
+ "processed_frames": [
1405
+ 25,
1406
+ 30,
1407
+ 35,
1408
+ 40,
1409
+ 45,
1410
+ 50,
1411
+ 55,
1412
+ 60,
1413
+ 65,
1414
+ 70,
1415
+ 75,
1416
+ 80,
1417
+ 85,
1418
+ 90,
1419
+ 95,
1420
+ 100,
1421
+ 105,
1422
+ 110,
1423
+ 115,
1424
+ 120,
1425
+ 125,
1426
+ 130,
1427
+ 135,
1428
+ 140,
1429
+ 145,
1430
+ 150,
1431
+ 155,
1432
+ 160,
1433
+ 165,
1434
+ 170,
1435
+ 175,
1436
+ 180,
1437
+ 185,
1438
+ 190,
1439
+ 195,
1440
+ 200,
1441
+ 205,
1442
+ 210,
1443
+ 215,
1444
+ 220,
1445
+ 225,
1446
+ 230,
1447
+ 235,
1448
+ 240,
1449
+ 245,
1450
+ 250,
1451
+ 255,
1452
+ 260,
1453
+ 265,
1454
+ 270,
1455
+ 275,
1456
+ 280,
1457
+ 285,
1458
+ 290,
1459
+ 295,
1460
+ 300,
1461
+ 305,
1462
+ 310,
1463
+ 315,
1464
+ 320,
1465
+ 325,
1466
+ 330,
1467
+ 335,
1468
+ 340,
1469
+ 345,
1470
+ 350,
1471
+ 355,
1472
+ 360,
1473
+ 365,
1474
+ 370,
1475
+ 375,
1476
+ 380,
1477
+ 385,
1478
+ 390,
1479
+ 395,
1480
+ 400,
1481
+ 405,
1482
+ 410,
1483
+ 415,
1484
+ 420,
1485
+ 425,
1486
+ 430,
1487
+ 435,
1488
+ 440,
1489
+ 445,
1490
+ 450,
1491
+ 455,
1492
+ 460,
1493
+ 465,
1494
+ 470,
1495
+ 475,
1496
+ 480,
1497
+ 485,
1498
+ 490,
1499
+ 495,
1500
+ 500,
1501
+ 505,
1502
+ 510,
1503
+ 515,
1504
+ 520,
1505
+ 525,
1506
+ 530,
1507
+ 535,
1508
+ 540,
1509
+ 545,
1510
+ 550,
1511
+ 555,
1512
+ 560,
1513
+ 565,
1514
+ 570
1515
+ ],
1516
+ "decoded_range": [
1517
+ 0,
1518
+ 572
1519
+ ],
1520
+ "clusters": [
1521
+ [
1522
+ [
1523
+ 180.16492165242153,
1524
+ 124.98789173789174
1525
+ ],
1526
+ 52
1527
+ ],
1528
+ [
1529
+ [
1530
+ 176.27777777777774,
1531
+ 192.83333333333334
1532
+ ],
1533
+ 27
1534
+ ]
1535
+ ],
1536
+ "selected": [],
1537
+ "inferred_grid": [
1538
+ [
1539
+ 180.16492165242153,
1540
+ 124.98789173789174
1541
+ ],
1542
+ [
1543
+ 176.27777777777774,
1544
+ 192.83333333333334
1545
+ ],
1546
+ [
1547
+ 108.43233618233614,
1548
+ 188.94618945868956
1549
+ ],
1550
+ [
1551
+ 112.31948005697993,
1552
+ 121.10074786324796
1553
+ ]
1554
+ ],
1555
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
1556
+ "observed_side_length": 67.95670557629069,
1557
+ "other_cluster_residuals": []
1558
+ },
1559
+ {
1560
+ "demo": "demo_09",
1561
+ "processed_frames": [
1562
+ 25,
1563
+ 30,
1564
+ 35,
1565
+ 40,
1566
+ 45,
1567
+ 50,
1568
+ 55,
1569
+ 60,
1570
+ 65,
1571
+ 70,
1572
+ 75,
1573
+ 80,
1574
+ 85,
1575
+ 90,
1576
+ 95,
1577
+ 100,
1578
+ 105,
1579
+ 110,
1580
+ 115,
1581
+ 120,
1582
+ 125,
1583
+ 130,
1584
+ 135,
1585
+ 140,
1586
+ 145,
1587
+ 150,
1588
+ 155,
1589
+ 160,
1590
+ 165,
1591
+ 170,
1592
+ 175,
1593
+ 180,
1594
+ 185,
1595
+ 190,
1596
+ 195,
1597
+ 200,
1598
+ 205,
1599
+ 210,
1600
+ 215,
1601
+ 220,
1602
+ 225,
1603
+ 230,
1604
+ 235,
1605
+ 240,
1606
+ 245,
1607
+ 250,
1608
+ 255,
1609
+ 260,
1610
+ 265,
1611
+ 270,
1612
+ 275,
1613
+ 280,
1614
+ 285,
1615
+ 290,
1616
+ 295,
1617
+ 300,
1618
+ 305,
1619
+ 310,
1620
+ 315,
1621
+ 320,
1622
+ 325,
1623
+ 330,
1624
+ 335,
1625
+ 340,
1626
+ 345,
1627
+ 350,
1628
+ 355,
1629
+ 360,
1630
+ 365,
1631
+ 370,
1632
+ 375,
1633
+ 380,
1634
+ 385,
1635
+ 390,
1636
+ 395,
1637
+ 400,
1638
+ 405,
1639
+ 410,
1640
+ 415,
1641
+ 420,
1642
+ 425,
1643
+ 430,
1644
+ 435,
1645
+ 440,
1646
+ 445,
1647
+ 450,
1648
+ 455,
1649
+ 460,
1650
+ 465,
1651
+ 470,
1652
+ 475,
1653
+ 480,
1654
+ 485,
1655
+ 490,
1656
+ 495,
1657
+ 500,
1658
+ 505,
1659
+ 510,
1660
+ 515,
1661
+ 520,
1662
+ 525,
1663
+ 530,
1664
+ 535,
1665
+ 540,
1666
+ 545,
1667
+ 550,
1668
+ 555,
1669
+ 560,
1670
+ 565,
1671
+ 570,
1672
+ 575,
1673
+ 580,
1674
+ 585,
1675
+ 590,
1676
+ 595,
1677
+ 600,
1678
+ 605,
1679
+ 610,
1680
+ 615,
1681
+ 620,
1682
+ 625,
1683
+ 630,
1684
+ 635,
1685
+ 640,
1686
+ 645,
1687
+ 650,
1688
+ 655,
1689
+ 660,
1690
+ 665,
1691
+ 670,
1692
+ 675,
1693
+ 680,
1694
+ 685,
1695
+ 690,
1696
+ 695,
1697
+ 700,
1698
+ 705,
1699
+ 710,
1700
+ 715
1701
+ ],
1702
+ "decoded_range": [
1703
+ 0,
1704
+ 717
1705
+ ],
1706
+ "clusters": [
1707
+ [
1708
+ [
1709
+ 116.66464821222597,
1710
+ 203.03316958441684
1711
+ ],
1712
+ 102
1713
+ ],
1714
+ [
1715
+ [
1716
+ 183.94444444444488,
1717
+ 190.94444444444488
1718
+ ],
1719
+ 95
1720
+ ]
1721
+ ],
1722
+ "selected": [],
1723
+ "inferred_grid": [
1724
+ [
1725
+ 116.66464821222597,
1726
+ 203.03316958441684
1727
+ ],
1728
+ [
1729
+ 183.94444444444488,
1730
+ 190.94444444444488
1731
+ ],
1732
+ [
1733
+ 171.85571930447293,
1734
+ 123.66464821222597
1735
+ ],
1736
+ [
1737
+ 104.57592307225401,
1738
+ 135.7533733521979
1739
+ ]
1740
+ ],
1741
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
1742
+ "observed_side_length": 68.35721071371101,
1743
+ "other_cluster_residuals": []
1744
+ },
1745
+ {
1746
+ "demo": "demo_10",
1747
+ "processed_frames": [
1748
+ 25,
1749
+ 30,
1750
+ 35,
1751
+ 40,
1752
+ 45,
1753
+ 50,
1754
+ 55,
1755
+ 60,
1756
+ 65,
1757
+ 70,
1758
+ 75,
1759
+ 80,
1760
+ 85,
1761
+ 90,
1762
+ 95,
1763
+ 100,
1764
+ 105,
1765
+ 110,
1766
+ 115,
1767
+ 120,
1768
+ 125,
1769
+ 130,
1770
+ 135,
1771
+ 140,
1772
+ 145,
1773
+ 150,
1774
+ 155,
1775
+ 160,
1776
+ 165,
1777
+ 170,
1778
+ 175,
1779
+ 180,
1780
+ 185,
1781
+ 190,
1782
+ 195,
1783
+ 200,
1784
+ 205,
1785
+ 210,
1786
+ 215,
1787
+ 220,
1788
+ 225,
1789
+ 230,
1790
+ 235,
1791
+ 240,
1792
+ 245,
1793
+ 250,
1794
+ 255,
1795
+ 260,
1796
+ 265,
1797
+ 270,
1798
+ 275,
1799
+ 280,
1800
+ 285,
1801
+ 290,
1802
+ 295,
1803
+ 300,
1804
+ 305,
1805
+ 310,
1806
+ 315,
1807
+ 320,
1808
+ 325,
1809
+ 330,
1810
+ 335,
1811
+ 340,
1812
+ 345,
1813
+ 350,
1814
+ 355,
1815
+ 360,
1816
+ 365,
1817
+ 370,
1818
+ 375,
1819
+ 380,
1820
+ 385,
1821
+ 390,
1822
+ 395,
1823
+ 400,
1824
+ 405,
1825
+ 410,
1826
+ 415,
1827
+ 420,
1828
+ 425,
1829
+ 430,
1830
+ 435,
1831
+ 440,
1832
+ 445,
1833
+ 450,
1834
+ 455,
1835
+ 460,
1836
+ 465,
1837
+ 470
1838
+ ],
1839
+ "decoded_range": [
1840
+ 0,
1841
+ 473
1842
+ ],
1843
+ "clusters": [
1844
+ [
1845
+ [
1846
+ 117.68816267247641,
1847
+ 120.9914306463326
1848
+ ],
1849
+ 45
1850
+ ],
1851
+ [
1852
+ [
1853
+ 104.58989626358046,
1854
+ 188.64160103772292
1855
+ ],
1856
+ 19
1857
+ ]
1858
+ ],
1859
+ "selected": [],
1860
+ "inferred_grid": [
1861
+ [
1862
+ 117.68816267247641,
1863
+ 120.9914306463326
1864
+ ],
1865
+ [
1866
+ 104.58989626358046,
1867
+ 188.64160103772292
1868
+ ],
1869
+ [
1870
+ 172.24006665497078,
1871
+ 201.73986744661886
1872
+ ],
1873
+ [
1874
+ 185.33833306386674,
1875
+ 134.08969705522856
1876
+ ]
1877
+ ],
1878
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
1879
+ "observed_side_length": 68.90653188851225,
1880
+ "other_cluster_residuals": []
1881
+ },
1882
+ {
1883
+ "demo": "demo_11",
1884
+ "processed_frames": [
1885
+ 25,
1886
+ 30,
1887
+ 35,
1888
+ 40,
1889
+ 45,
1890
+ 50,
1891
+ 55,
1892
+ 60,
1893
+ 65,
1894
+ 70,
1895
+ 75,
1896
+ 80,
1897
+ 85,
1898
+ 90,
1899
+ 95,
1900
+ 100,
1901
+ 105,
1902
+ 110,
1903
+ 115,
1904
+ 120,
1905
+ 125,
1906
+ 130,
1907
+ 135,
1908
+ 140,
1909
+ 145,
1910
+ 150,
1911
+ 155,
1912
+ 160,
1913
+ 165,
1914
+ 170,
1915
+ 175,
1916
+ 180,
1917
+ 185,
1918
+ 190,
1919
+ 195,
1920
+ 200,
1921
+ 205,
1922
+ 210,
1923
+ 215,
1924
+ 220,
1925
+ 225,
1926
+ 230,
1927
+ 235,
1928
+ 240,
1929
+ 245,
1930
+ 250,
1931
+ 255,
1932
+ 260,
1933
+ 265,
1934
+ 270,
1935
+ 275,
1936
+ 280,
1937
+ 285,
1938
+ 290,
1939
+ 295,
1940
+ 300,
1941
+ 305,
1942
+ 310,
1943
+ 315,
1944
+ 320,
1945
+ 325,
1946
+ 330,
1947
+ 335,
1948
+ 340,
1949
+ 345,
1950
+ 350,
1951
+ 355,
1952
+ 360,
1953
+ 365,
1954
+ 370,
1955
+ 375,
1956
+ 380,
1957
+ 385,
1958
+ 390,
1959
+ 395,
1960
+ 400,
1961
+ 405,
1962
+ 410,
1963
+ 415,
1964
+ 420,
1965
+ 425,
1966
+ 430,
1967
+ 435,
1968
+ 440,
1969
+ 445,
1970
+ 450,
1971
+ 455,
1972
+ 460,
1973
+ 465,
1974
+ 470,
1975
+ 475,
1976
+ 480,
1977
+ 485,
1978
+ 490,
1979
+ 495,
1980
+ 500,
1981
+ 505,
1982
+ 510,
1983
+ 515,
1984
+ 520,
1985
+ 525,
1986
+ 530,
1987
+ 535,
1988
+ 540,
1989
+ 545,
1990
+ 550,
1991
+ 555,
1992
+ 560,
1993
+ 565,
1994
+ 570,
1995
+ 575,
1996
+ 580,
1997
+ 585,
1998
+ 590,
1999
+ 595,
2000
+ 600,
2001
+ 605,
2002
+ 610,
2003
+ 615,
2004
+ 620,
2005
+ 625,
2006
+ 630,
2007
+ 635,
2008
+ 640,
2009
+ 645,
2010
+ 650,
2011
+ 655,
2012
+ 660,
2013
+ 665,
2014
+ 670,
2015
+ 675,
2016
+ 680,
2017
+ 685,
2018
+ 690,
2019
+ 695,
2020
+ 700,
2021
+ 705,
2022
+ 710,
2023
+ 715,
2024
+ 720,
2025
+ 725,
2026
+ 730,
2027
+ 735,
2028
+ 740,
2029
+ 745,
2030
+ 750,
2031
+ 755,
2032
+ 760,
2033
+ 765,
2034
+ 770,
2035
+ 775,
2036
+ 780,
2037
+ 785,
2038
+ 790,
2039
+ 795,
2040
+ 800,
2041
+ 805,
2042
+ 810,
2043
+ 815,
2044
+ 820,
2045
+ 825,
2046
+ 830,
2047
+ 835,
2048
+ 840,
2049
+ 845,
2050
+ 850,
2051
+ 855,
2052
+ 860,
2053
+ 865,
2054
+ 870,
2055
+ 875,
2056
+ 880,
2057
+ 885,
2058
+ 890,
2059
+ 895,
2060
+ 900
2061
+ ],
2062
+ "decoded_range": [
2063
+ 0,
2064
+ 902
2065
+ ],
2066
+ "clusters": [
2067
+ [
2068
+ [
2069
+ 178.8016086117347,
2070
+ 191.59564298804779
2071
+ ],
2072
+ 158
2073
+ ],
2074
+ [
2075
+ [
2076
+ 176.54385964912265,
2077
+ 123.19298245614029
2078
+ ],
2079
+ 67
2080
+ ],
2081
+ [
2082
+ [
2083
+ 110.4920634920635,
2084
+ 193.5079365079365
2085
+ ],
2086
+ 35
2087
+ ]
2088
+ ],
2089
+ "selected": [],
2090
+ "inferred_grid": [
2091
+ [
2092
+ 178.8016086117347,
2093
+ 191.59564298804779
2094
+ ],
2095
+ [
2096
+ 176.54385964912265,
2097
+ 123.19298245614029
2098
+ ],
2099
+ [
2100
+ 108.14119911721515,
2101
+ 125.45073141875234
2102
+ ],
2103
+ [
2104
+ 110.3989480798272,
2105
+ 193.85339195065984
2106
+ ]
2107
+ ],
2108
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
2109
+ "observed_side_length": 68.43991085778497,
2110
+ "other_cluster_residuals": [
2111
+ 0.3577847717596556
2112
+ ]
2113
+ },
2114
+ {
2115
+ "demo": "demo_12",
2116
+ "processed_frames": [
2117
+ 25,
2118
+ 30,
2119
+ 35,
2120
+ 40,
2121
+ 45,
2122
+ 50,
2123
+ 55,
2124
+ 60,
2125
+ 65,
2126
+ 70,
2127
+ 75,
2128
+ 80,
2129
+ 85,
2130
+ 90,
2131
+ 95,
2132
+ 100,
2133
+ 105,
2134
+ 110,
2135
+ 115,
2136
+ 120,
2137
+ 125,
2138
+ 130,
2139
+ 135,
2140
+ 140,
2141
+ 145,
2142
+ 150,
2143
+ 155,
2144
+ 160,
2145
+ 165,
2146
+ 170,
2147
+ 175,
2148
+ 180,
2149
+ 185,
2150
+ 190,
2151
+ 195,
2152
+ 200,
2153
+ 205,
2154
+ 210,
2155
+ 215,
2156
+ 220,
2157
+ 225,
2158
+ 230,
2159
+ 235,
2160
+ 240,
2161
+ 245,
2162
+ 250,
2163
+ 255,
2164
+ 260,
2165
+ 265,
2166
+ 270,
2167
+ 275,
2168
+ 280,
2169
+ 285,
2170
+ 290,
2171
+ 295,
2172
+ 300,
2173
+ 305,
2174
+ 310,
2175
+ 315,
2176
+ 320,
2177
+ 325,
2178
+ 330,
2179
+ 335,
2180
+ 340,
2181
+ 345,
2182
+ 350,
2183
+ 355,
2184
+ 360,
2185
+ 365,
2186
+ 370,
2187
+ 375,
2188
+ 380,
2189
+ 385,
2190
+ 390,
2191
+ 395,
2192
+ 400,
2193
+ 405,
2194
+ 410,
2195
+ 415,
2196
+ 420,
2197
+ 425,
2198
+ 430,
2199
+ 435,
2200
+ 440,
2201
+ 445,
2202
+ 450,
2203
+ 455,
2204
+ 460,
2205
+ 465,
2206
+ 470,
2207
+ 475,
2208
+ 480,
2209
+ 485,
2210
+ 490,
2211
+ 495,
2212
+ 500,
2213
+ 505,
2214
+ 510,
2215
+ 515,
2216
+ 520,
2217
+ 525,
2218
+ 530,
2219
+ 535,
2220
+ 540,
2221
+ 545,
2222
+ 550,
2223
+ 555,
2224
+ 560,
2225
+ 565,
2226
+ 570,
2227
+ 575,
2228
+ 580,
2229
+ 585,
2230
+ 590,
2231
+ 595,
2232
+ 600,
2233
+ 605,
2234
+ 610,
2235
+ 615,
2236
+ 620,
2237
+ 625,
2238
+ 630,
2239
+ 635,
2240
+ 640,
2241
+ 645,
2242
+ 650,
2243
+ 655,
2244
+ 660,
2245
+ 665,
2246
+ 670,
2247
+ 675,
2248
+ 680,
2249
+ 685,
2250
+ 690,
2251
+ 695,
2252
+ 700,
2253
+ 705,
2254
+ 710,
2255
+ 715,
2256
+ 720,
2257
+ 725,
2258
+ 730,
2259
+ 735,
2260
+ 740,
2261
+ 745,
2262
+ 750,
2263
+ 755,
2264
+ 760,
2265
+ 765,
2266
+ 770,
2267
+ 775,
2268
+ 780,
2269
+ 785,
2270
+ 790,
2271
+ 795,
2272
+ 800,
2273
+ 805,
2274
+ 810,
2275
+ 815,
2276
+ 820,
2277
+ 825,
2278
+ 830,
2279
+ 835,
2280
+ 840,
2281
+ 845,
2282
+ 850,
2283
+ 855,
2284
+ 860,
2285
+ 865,
2286
+ 870,
2287
+ 875,
2288
+ 880,
2289
+ 885,
2290
+ 890,
2291
+ 895,
2292
+ 900,
2293
+ 905,
2294
+ 910,
2295
+ 915,
2296
+ 920,
2297
+ 925,
2298
+ 930,
2299
+ 935,
2300
+ 940
2301
+ ],
2302
+ "decoded_range": [
2303
+ 0,
2304
+ 941
2305
+ ],
2306
+ "clusters": [
2307
+ [
2308
+ [
2309
+ 175.71180246365446,
2310
+ 125.63670595522439
2311
+ ],
2312
+ 81
2313
+ ],
2314
+ [
2315
+ [
2316
+ 107.57625001137849,
2317
+ 120.98125065149327
2318
+ ],
2319
+ 43
2320
+ ]
2321
+ ],
2322
+ "selected": [],
2323
+ "inferred_grid": [
2324
+ [
2325
+ 175.71180246365446,
2326
+ 125.63670595522439
2327
+ ],
2328
+ [
2329
+ 107.57625001137849,
2330
+ 120.98125065149327
2331
+ ],
2332
+ [
2333
+ 102.92079470764737,
2334
+ 189.11680310376926
2335
+ ],
2336
+ [
2337
+ 171.05634715992335,
2338
+ 193.77225840750037
2339
+ ]
2340
+ ],
2341
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
2342
+ "observed_side_length": 68.2944124512532,
2343
+ "other_cluster_residuals": []
2344
+ },
2345
+ {
2346
+ "demo": "demo_13",
2347
+ "processed_frames": [
2348
+ 25,
2349
+ 30,
2350
+ 35,
2351
+ 40,
2352
+ 45,
2353
+ 50,
2354
+ 55,
2355
+ 60,
2356
+ 65,
2357
+ 70,
2358
+ 75,
2359
+ 80,
2360
+ 85,
2361
+ 90,
2362
+ 95,
2363
+ 100,
2364
+ 105,
2365
+ 110,
2366
+ 115,
2367
+ 120,
2368
+ 125,
2369
+ 130,
2370
+ 135,
2371
+ 140,
2372
+ 145,
2373
+ 150,
2374
+ 155,
2375
+ 160,
2376
+ 165,
2377
+ 170,
2378
+ 175,
2379
+ 180,
2380
+ 185,
2381
+ 190,
2382
+ 195,
2383
+ 200,
2384
+ 205,
2385
+ 210,
2386
+ 215,
2387
+ 220,
2388
+ 225,
2389
+ 230,
2390
+ 235,
2391
+ 240,
2392
+ 245,
2393
+ 250,
2394
+ 255,
2395
+ 260,
2396
+ 265,
2397
+ 270,
2398
+ 275,
2399
+ 280,
2400
+ 285,
2401
+ 290,
2402
+ 295,
2403
+ 300,
2404
+ 305,
2405
+ 310,
2406
+ 315,
2407
+ 320,
2408
+ 325,
2409
+ 330,
2410
+ 335,
2411
+ 340,
2412
+ 345,
2413
+ 350,
2414
+ 355,
2415
+ 360,
2416
+ 365,
2417
+ 370,
2418
+ 375,
2419
+ 380,
2420
+ 385,
2421
+ 390,
2422
+ 395,
2423
+ 400,
2424
+ 405,
2425
+ 410,
2426
+ 415,
2427
+ 420,
2428
+ 425,
2429
+ 430,
2430
+ 435,
2431
+ 440,
2432
+ 445,
2433
+ 450,
2434
+ 455,
2435
+ 460,
2436
+ 465,
2437
+ 470,
2438
+ 475,
2439
+ 480,
2440
+ 485,
2441
+ 490,
2442
+ 495,
2443
+ 500,
2444
+ 505,
2445
+ 510,
2446
+ 515,
2447
+ 520,
2448
+ 525,
2449
+ 530,
2450
+ 535,
2451
+ 540,
2452
+ 545,
2453
+ 550,
2454
+ 555,
2455
+ 560,
2456
+ 565,
2457
+ 570,
2458
+ 575,
2459
+ 580,
2460
+ 585,
2461
+ 590,
2462
+ 595,
2463
+ 600,
2464
+ 605,
2465
+ 610,
2466
+ 615,
2467
+ 620,
2468
+ 625,
2469
+ 630,
2470
+ 635,
2471
+ 640,
2472
+ 645,
2473
+ 650,
2474
+ 655,
2475
+ 660,
2476
+ 665,
2477
+ 670,
2478
+ 675,
2479
+ 680,
2480
+ 685,
2481
+ 690,
2482
+ 695,
2483
+ 700,
2484
+ 705,
2485
+ 710,
2486
+ 715,
2487
+ 720,
2488
+ 725,
2489
+ 730,
2490
+ 735,
2491
+ 740,
2492
+ 745,
2493
+ 750,
2494
+ 755,
2495
+ 760,
2496
+ 765,
2497
+ 770,
2498
+ 775,
2499
+ 780,
2500
+ 785,
2501
+ 790,
2502
+ 795,
2503
+ 800,
2504
+ 805,
2505
+ 810,
2506
+ 815,
2507
+ 820,
2508
+ 825,
2509
+ 830,
2510
+ 835,
2511
+ 840,
2512
+ 845,
2513
+ 850,
2514
+ 855,
2515
+ 860,
2516
+ 865,
2517
+ 870,
2518
+ 875,
2519
+ 880
2520
+ ],
2521
+ "decoded_range": [
2522
+ 0,
2523
+ 880
2524
+ ],
2525
+ "clusters": [
2526
+ [
2527
+ [
2528
+ 178.2753454947482,
2529
+ 196.84002211166364
2530
+ ],
2531
+ 134
2532
+ ],
2533
+ [
2534
+ [
2535
+ 181.90196078431364,
2536
+ 128.686274509804
2537
+ ],
2538
+ 52
2539
+ ],
2540
+ [
2541
+ [
2542
+ 109.55555555555557,
2543
+ 193.36881466486733
2544
+ ],
2545
+ 18
2546
+ ]
2547
+ ],
2548
+ "selected": [],
2549
+ "inferred_grid": [
2550
+ [
2551
+ 178.2753454947482,
2552
+ 196.84002211166364
2553
+ ],
2554
+ [
2555
+ 181.90196078431364,
2556
+ 128.686274509804
2557
+ ],
2558
+ [
2559
+ 113.74821318245401,
2560
+ 125.05965922023856
2561
+ ],
2562
+ [
2563
+ 110.12159789288856,
2564
+ 193.2134068220982
2565
+ ]
2566
+ ],
2567
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
2568
+ "observed_side_length": 68.25016960152186,
2569
+ "other_cluster_residuals": [
2570
+ 0.5869885222451531
2571
+ ]
2572
+ },
2573
+ {
2574
+ "demo": "demo_14",
2575
+ "processed_frames": [
2576
+ 25,
2577
+ 30,
2578
+ 35,
2579
+ 40,
2580
+ 45,
2581
+ 50,
2582
+ 55,
2583
+ 60,
2584
+ 65,
2585
+ 70,
2586
+ 75,
2587
+ 80,
2588
+ 85,
2589
+ 90,
2590
+ 95,
2591
+ 100,
2592
+ 105,
2593
+ 110,
2594
+ 115,
2595
+ 120,
2596
+ 125,
2597
+ 130,
2598
+ 135,
2599
+ 140,
2600
+ 145,
2601
+ 150,
2602
+ 155,
2603
+ 160,
2604
+ 165,
2605
+ 170,
2606
+ 175,
2607
+ 180,
2608
+ 185,
2609
+ 190,
2610
+ 195,
2611
+ 200,
2612
+ 205,
2613
+ 210,
2614
+ 215,
2615
+ 220,
2616
+ 225,
2617
+ 230,
2618
+ 235,
2619
+ 240,
2620
+ 245,
2621
+ 250,
2622
+ 255,
2623
+ 260,
2624
+ 265,
2625
+ 270,
2626
+ 275,
2627
+ 280,
2628
+ 285,
2629
+ 290,
2630
+ 295,
2631
+ 300,
2632
+ 305,
2633
+ 310,
2634
+ 315,
2635
+ 320,
2636
+ 325,
2637
+ 330,
2638
+ 335,
2639
+ 340,
2640
+ 345,
2641
+ 350,
2642
+ 355,
2643
+ 360,
2644
+ 365,
2645
+ 370,
2646
+ 375,
2647
+ 380,
2648
+ 385,
2649
+ 390,
2650
+ 395,
2651
+ 400,
2652
+ 405,
2653
+ 410,
2654
+ 415,
2655
+ 420,
2656
+ 425,
2657
+ 430,
2658
+ 435,
2659
+ 440,
2660
+ 445,
2661
+ 450,
2662
+ 455,
2663
+ 460,
2664
+ 465,
2665
+ 470,
2666
+ 475,
2667
+ 480,
2668
+ 485,
2669
+ 490,
2670
+ 495,
2671
+ 500,
2672
+ 505,
2673
+ 510,
2674
+ 515,
2675
+ 520,
2676
+ 525,
2677
+ 530,
2678
+ 535,
2679
+ 540,
2680
+ 545,
2681
+ 550,
2682
+ 555,
2683
+ 560,
2684
+ 565,
2685
+ 570,
2686
+ 575,
2687
+ 580,
2688
+ 585,
2689
+ 590,
2690
+ 595,
2691
+ 600
2692
+ ],
2693
+ "decoded_range": [
2694
+ 0,
2695
+ 602
2696
+ ],
2697
+ "clusters": [
2698
+ [
2699
+ [
2700
+ 108.5377087921825,
2701
+ 199.2458733112243
2702
+ ],
2703
+ 60
2704
+ ],
2705
+ [
2706
+ [
2707
+ 176.75,
2708
+ 194.125
2709
+ ],
2710
+ 28
2711
+ ],
2712
+ [
2713
+ [
2714
+ 90.97458017023234,
2715
+ 116.41039797561535
2716
+ ],
2717
+ 3
2718
+ ]
2719
+ ],
2720
+ "selected": [],
2721
+ "inferred_grid": [
2722
+ [
2723
+ 108.5377087921825,
2724
+ 199.2458733112243
2725
+ ],
2726
+ [
2727
+ 176.75,
2728
+ 194.125
2729
+ ],
2730
+ [
2731
+ 171.6291266887757,
2732
+ 125.9127087921825
2733
+ ],
2734
+ [
2735
+ 103.4168354809582,
2736
+ 131.0335821034068
2737
+ ]
2738
+ ],
2739
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
2740
+ "observed_side_length": 68.40423974644924,
2741
+ "other_cluster_residuals": [
2742
+ 19.20018831294571
2743
+ ]
2744
+ },
2745
+ {
2746
+ "demo": "demo_15",
2747
+ "processed_frames": [
2748
+ 25,
2749
+ 30,
2750
+ 35,
2751
+ 40,
2752
+ 45,
2753
+ 50,
2754
+ 55,
2755
+ 60,
2756
+ 65,
2757
+ 70,
2758
+ 75,
2759
+ 80,
2760
+ 85,
2761
+ 90,
2762
+ 95,
2763
+ 100,
2764
+ 105,
2765
+ 110,
2766
+ 115,
2767
+ 120,
2768
+ 125,
2769
+ 130,
2770
+ 135,
2771
+ 140,
2772
+ 145,
2773
+ 150,
2774
+ 155,
2775
+ 160,
2776
+ 165,
2777
+ 170,
2778
+ 175,
2779
+ 180,
2780
+ 185,
2781
+ 190,
2782
+ 195,
2783
+ 200,
2784
+ 205,
2785
+ 210,
2786
+ 215,
2787
+ 220,
2788
+ 225,
2789
+ 230,
2790
+ 235,
2791
+ 240,
2792
+ 245,
2793
+ 250,
2794
+ 255,
2795
+ 260,
2796
+ 265,
2797
+ 270,
2798
+ 275,
2799
+ 280,
2800
+ 285,
2801
+ 290,
2802
+ 295,
2803
+ 300,
2804
+ 305,
2805
+ 310,
2806
+ 315,
2807
+ 320,
2808
+ 325,
2809
+ 330,
2810
+ 335,
2811
+ 340,
2812
+ 345,
2813
+ 350,
2814
+ 355,
2815
+ 360,
2816
+ 365,
2817
+ 370,
2818
+ 375,
2819
+ 380,
2820
+ 385,
2821
+ 390,
2822
+ 395,
2823
+ 400,
2824
+ 405,
2825
+ 410,
2826
+ 415,
2827
+ 420,
2828
+ 425,
2829
+ 430,
2830
+ 435,
2831
+ 440,
2832
+ 445,
2833
+ 450,
2834
+ 455,
2835
+ 460,
2836
+ 465,
2837
+ 470,
2838
+ 475,
2839
+ 480,
2840
+ 485,
2841
+ 490,
2842
+ 495,
2843
+ 500,
2844
+ 505,
2845
+ 510,
2846
+ 515,
2847
+ 520,
2848
+ 525,
2849
+ 530,
2850
+ 535,
2851
+ 540,
2852
+ 545,
2853
+ 550,
2854
+ 555,
2855
+ 560,
2856
+ 565,
2857
+ 570,
2858
+ 575,
2859
+ 580,
2860
+ 585,
2861
+ 590,
2862
+ 595,
2863
+ 600,
2864
+ 605,
2865
+ 610,
2866
+ 615,
2867
+ 620,
2868
+ 625,
2869
+ 630,
2870
+ 635,
2871
+ 640,
2872
+ 645,
2873
+ 650,
2874
+ 655,
2875
+ 660,
2876
+ 665,
2877
+ 670,
2878
+ 675,
2879
+ 680,
2880
+ 685,
2881
+ 690,
2882
+ 695,
2883
+ 700,
2884
+ 705,
2885
+ 710,
2886
+ 715,
2887
+ 720
2888
+ ],
2889
+ "decoded_range": [
2890
+ 0,
2891
+ 722
2892
+ ],
2893
+ "clusters": [
2894
+ [
2895
+ [
2896
+ 106.58721174004206,
2897
+ 188.20461215932923
2898
+ ],
2899
+ 106
2900
+ ],
2901
+ [
2902
+ [
2903
+ 174.6111111111112,
2904
+ 195.111111111111
2905
+ ],
2906
+ 96
2907
+ ]
2908
+ ],
2909
+ "selected": [],
2910
+ "inferred_grid": [
2911
+ [
2912
+ 106.58721174004206,
2913
+ 188.20461215932923
2914
+ ],
2915
+ [
2916
+ 174.6111111111112,
2917
+ 195.111111111111
2918
+ ],
2919
+ [
2920
+ 181.51761006289297,
2921
+ 127.08721174004187
2922
+ ],
2923
+ [
2924
+ 113.49371069182384,
2925
+ 120.1807127882601
2926
+ ]
2927
+ ],
2928
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
2929
+ "observed_side_length": 68.37361050446512,
2930
+ "other_cluster_residuals": []
2931
+ },
2932
+ {
2933
+ "demo": "demo_16",
2934
+ "processed_frames": [
2935
+ 25,
2936
+ 30,
2937
+ 35,
2938
+ 40,
2939
+ 45,
2940
+ 50,
2941
+ 55,
2942
+ 60,
2943
+ 65,
2944
+ 70,
2945
+ 75,
2946
+ 80,
2947
+ 85,
2948
+ 90,
2949
+ 95,
2950
+ 100,
2951
+ 105,
2952
+ 110,
2953
+ 115,
2954
+ 120,
2955
+ 125,
2956
+ 130,
2957
+ 135,
2958
+ 140,
2959
+ 145,
2960
+ 150,
2961
+ 155,
2962
+ 160,
2963
+ 165,
2964
+ 170,
2965
+ 175,
2966
+ 180,
2967
+ 185,
2968
+ 190,
2969
+ 195,
2970
+ 200,
2971
+ 205,
2972
+ 210,
2973
+ 215,
2974
+ 220,
2975
+ 225,
2976
+ 230,
2977
+ 235,
2978
+ 240,
2979
+ 245,
2980
+ 250,
2981
+ 255,
2982
+ 260,
2983
+ 265,
2984
+ 270,
2985
+ 275,
2986
+ 280,
2987
+ 285,
2988
+ 290,
2989
+ 295,
2990
+ 300,
2991
+ 305,
2992
+ 310,
2993
+ 315,
2994
+ 320,
2995
+ 325,
2996
+ 330,
2997
+ 335,
2998
+ 340,
2999
+ 345,
3000
+ 350,
3001
+ 355,
3002
+ 360,
3003
+ 365,
3004
+ 370,
3005
+ 375,
3006
+ 380,
3007
+ 385,
3008
+ 390,
3009
+ 395,
3010
+ 400,
3011
+ 405,
3012
+ 410,
3013
+ 415,
3014
+ 420,
3015
+ 425,
3016
+ 430,
3017
+ 435,
3018
+ 440,
3019
+ 445,
3020
+ 450,
3021
+ 455,
3022
+ 460,
3023
+ 465,
3024
+ 470,
3025
+ 475,
3026
+ 480,
3027
+ 485,
3028
+ 490,
3029
+ 495,
3030
+ 500,
3031
+ 505,
3032
+ 510,
3033
+ 515,
3034
+ 520,
3035
+ 525,
3036
+ 530,
3037
+ 535,
3038
+ 540,
3039
+ 545,
3040
+ 550,
3041
+ 555,
3042
+ 560,
3043
+ 565,
3044
+ 570,
3045
+ 575,
3046
+ 580,
3047
+ 585,
3048
+ 590,
3049
+ 595,
3050
+ 600,
3051
+ 605,
3052
+ 610,
3053
+ 615,
3054
+ 620,
3055
+ 625,
3056
+ 630,
3057
+ 635,
3058
+ 640,
3059
+ 645,
3060
+ 650,
3061
+ 655,
3062
+ 660,
3063
+ 665,
3064
+ 670,
3065
+ 675,
3066
+ 680,
3067
+ 685,
3068
+ 690,
3069
+ 695,
3070
+ 700,
3071
+ 705,
3072
+ 710,
3073
+ 715,
3074
+ 720,
3075
+ 725,
3076
+ 730,
3077
+ 735,
3078
+ 740,
3079
+ 745,
3080
+ 750,
3081
+ 755,
3082
+ 760,
3083
+ 765,
3084
+ 770,
3085
+ 775,
3086
+ 780,
3087
+ 785,
3088
+ 790,
3089
+ 795,
3090
+ 800,
3091
+ 805,
3092
+ 810,
3093
+ 815,
3094
+ 820,
3095
+ 825,
3096
+ 830,
3097
+ 835,
3098
+ 840,
3099
+ 845,
3100
+ 850,
3101
+ 855,
3102
+ 860,
3103
+ 865,
3104
+ 870,
3105
+ 875,
3106
+ 880,
3107
+ 885,
3108
+ 890,
3109
+ 895,
3110
+ 900,
3111
+ 905,
3112
+ 910,
3113
+ 915,
3114
+ 920,
3115
+ 925,
3116
+ 930,
3117
+ 935,
3118
+ 940,
3119
+ 945,
3120
+ 950
3121
+ ],
3122
+ "decoded_range": [
3123
+ 0,
3124
+ 953
3125
+ ],
3126
+ "clusters": [
3127
+ [
3128
+ [
3129
+ 187.5625,
3130
+ 186.9375
3131
+ ],
3132
+ 62
3133
+ ],
3134
+ [
3135
+ [
3136
+ 121.5217773300876,
3137
+ 205.26271020453862
3138
+ ],
3139
+ 57
3140
+ ],
3141
+ [
3142
+ [
3143
+ 89.63572287742767,
3144
+ 103.84088711419946
3145
+ ],
3146
+ 12
3147
+ ]
3148
+ ],
3149
+ "selected": [],
3150
+ "inferred_grid": [
3151
+ [
3152
+ 187.5625,
3153
+ 186.9375
3154
+ ],
3155
+ [
3156
+ 121.5217773300876,
3157
+ 205.26271020453862
3158
+ ],
3159
+ [
3160
+ 103.19656712554898,
3161
+ 139.22198753462624
3162
+ ],
3163
+ [
3164
+ 169.23728979546138,
3165
+ 120.8967773300876
3166
+ ]
3167
+ ],
3168
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
3169
+ "observed_side_length": 68.53605167942496,
3170
+ "other_cluster_residuals": [
3171
+ 37.89087969000098
3172
+ ]
3173
+ },
3174
+ {
3175
+ "demo": "demo_17",
3176
+ "processed_frames": [
3177
+ 25,
3178
+ 30,
3179
+ 35,
3180
+ 40,
3181
+ 45,
3182
+ 50,
3183
+ 55,
3184
+ 60,
3185
+ 65,
3186
+ 70,
3187
+ 75,
3188
+ 80,
3189
+ 85,
3190
+ 90,
3191
+ 95,
3192
+ 100,
3193
+ 105,
3194
+ 110,
3195
+ 115,
3196
+ 120,
3197
+ 125,
3198
+ 130,
3199
+ 135,
3200
+ 140,
3201
+ 145,
3202
+ 150,
3203
+ 155,
3204
+ 160,
3205
+ 165,
3206
+ 170,
3207
+ 175,
3208
+ 180,
3209
+ 185,
3210
+ 190,
3211
+ 195,
3212
+ 200,
3213
+ 205,
3214
+ 210,
3215
+ 215,
3216
+ 220,
3217
+ 225,
3218
+ 230,
3219
+ 235,
3220
+ 240,
3221
+ 245,
3222
+ 250,
3223
+ 255,
3224
+ 260,
3225
+ 265,
3226
+ 270,
3227
+ 275,
3228
+ 280,
3229
+ 285,
3230
+ 290,
3231
+ 295,
3232
+ 300,
3233
+ 305,
3234
+ 310,
3235
+ 315,
3236
+ 320,
3237
+ 325,
3238
+ 330,
3239
+ 335,
3240
+ 340,
3241
+ 345,
3242
+ 350,
3243
+ 355,
3244
+ 360,
3245
+ 365,
3246
+ 370,
3247
+ 375,
3248
+ 380,
3249
+ 385,
3250
+ 390,
3251
+ 395,
3252
+ 400,
3253
+ 405,
3254
+ 410,
3255
+ 415,
3256
+ 420,
3257
+ 425,
3258
+ 430,
3259
+ 435,
3260
+ 440,
3261
+ 445,
3262
+ 450,
3263
+ 455,
3264
+ 460,
3265
+ 465,
3266
+ 470,
3267
+ 475,
3268
+ 480,
3269
+ 485,
3270
+ 490,
3271
+ 495,
3272
+ 500,
3273
+ 505,
3274
+ 510,
3275
+ 515,
3276
+ 520,
3277
+ 525,
3278
+ 530,
3279
+ 535,
3280
+ 540,
3281
+ 545,
3282
+ 550,
3283
+ 555,
3284
+ 560,
3285
+ 565,
3286
+ 570,
3287
+ 575,
3288
+ 580,
3289
+ 585,
3290
+ 590,
3291
+ 595,
3292
+ 600,
3293
+ 605,
3294
+ 610,
3295
+ 615,
3296
+ 620,
3297
+ 625,
3298
+ 630,
3299
+ 635,
3300
+ 640,
3301
+ 645,
3302
+ 650,
3303
+ 655,
3304
+ 660,
3305
+ 665,
3306
+ 670,
3307
+ 675,
3308
+ 680,
3309
+ 685,
3310
+ 690,
3311
+ 695,
3312
+ 700,
3313
+ 705,
3314
+ 710,
3315
+ 715,
3316
+ 720,
3317
+ 725,
3318
+ 730,
3319
+ 735,
3320
+ 740,
3321
+ 745,
3322
+ 750,
3323
+ 755,
3324
+ 760,
3325
+ 765,
3326
+ 770,
3327
+ 775,
3328
+ 780,
3329
+ 785,
3330
+ 790,
3331
+ 795,
3332
+ 800,
3333
+ 805,
3334
+ 810,
3335
+ 815,
3336
+ 820,
3337
+ 825,
3338
+ 830,
3339
+ 835,
3340
+ 840,
3341
+ 845,
3342
+ 850,
3343
+ 855,
3344
+ 860,
3345
+ 865,
3346
+ 870,
3347
+ 875,
3348
+ 880,
3349
+ 885,
3350
+ 890,
3351
+ 895,
3352
+ 900,
3353
+ 905,
3354
+ 910,
3355
+ 915,
3356
+ 920,
3357
+ 925,
3358
+ 930,
3359
+ 935,
3360
+ 940,
3361
+ 945,
3362
+ 950,
3363
+ 955,
3364
+ 960,
3365
+ 965,
3366
+ 970,
3367
+ 975,
3368
+ 980
3369
+ ],
3370
+ "decoded_range": [
3371
+ 0,
3372
+ 981
3373
+ ],
3374
+ "clusters": [
3375
+ [
3376
+ [
3377
+ 100.73550606481628,
3378
+ 189.67205502033067
3379
+ ],
3380
+ 87
3381
+ ],
3382
+ [
3383
+ [
3384
+ 114.11041666666667,
3385
+ 122.75393518518517
3386
+ ],
3387
+ 16
3388
+ ]
3389
+ ],
3390
+ "selected": [],
3391
+ "inferred_grid": [
3392
+ [
3393
+ 100.73550606481628,
3394
+ 189.67205502033067
3395
+ ],
3396
+ [
3397
+ 114.11041666666667,
3398
+ 122.75393518518517
3399
+ ],
3400
+ [
3401
+ 181.02853650181217,
3402
+ 136.12884578703557
3403
+ ],
3404
+ [
3405
+ 167.65362589996178,
3406
+ 203.04696562218106
3407
+ ]
3408
+ ],
3409
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
3410
+ "observed_side_length": 68.24165147384979,
3411
+ "other_cluster_residuals": []
3412
+ },
3413
+ {
3414
+ "demo": "demo_18",
3415
+ "processed_frames": [
3416
+ 25,
3417
+ 30,
3418
+ 35,
3419
+ 40,
3420
+ 45,
3421
+ 50,
3422
+ 55,
3423
+ 60,
3424
+ 65,
3425
+ 70,
3426
+ 75,
3427
+ 80,
3428
+ 85,
3429
+ 90,
3430
+ 95,
3431
+ 100,
3432
+ 105,
3433
+ 110,
3434
+ 115,
3435
+ 120,
3436
+ 125,
3437
+ 130,
3438
+ 135,
3439
+ 140,
3440
+ 145,
3441
+ 150,
3442
+ 155,
3443
+ 160,
3444
+ 165,
3445
+ 170,
3446
+ 175,
3447
+ 180,
3448
+ 185,
3449
+ 190,
3450
+ 195,
3451
+ 200,
3452
+ 205,
3453
+ 210,
3454
+ 215,
3455
+ 220,
3456
+ 225,
3457
+ 230,
3458
+ 235,
3459
+ 240,
3460
+ 245,
3461
+ 250,
3462
+ 255,
3463
+ 260,
3464
+ 265,
3465
+ 270,
3466
+ 275,
3467
+ 280,
3468
+ 285,
3469
+ 290,
3470
+ 295,
3471
+ 300,
3472
+ 305,
3473
+ 310,
3474
+ 315,
3475
+ 320,
3476
+ 325,
3477
+ 330,
3478
+ 335,
3479
+ 340,
3480
+ 345,
3481
+ 350,
3482
+ 355,
3483
+ 360,
3484
+ 365,
3485
+ 370,
3486
+ 375,
3487
+ 380,
3488
+ 385,
3489
+ 390,
3490
+ 395,
3491
+ 400,
3492
+ 405,
3493
+ 410,
3494
+ 415,
3495
+ 420,
3496
+ 425,
3497
+ 430,
3498
+ 435,
3499
+ 440,
3500
+ 445,
3501
+ 450,
3502
+ 455,
3503
+ 460,
3504
+ 465,
3505
+ 470,
3506
+ 475,
3507
+ 480,
3508
+ 485,
3509
+ 490,
3510
+ 495,
3511
+ 500,
3512
+ 505,
3513
+ 510,
3514
+ 515,
3515
+ 520,
3516
+ 525,
3517
+ 530,
3518
+ 535,
3519
+ 540,
3520
+ 545,
3521
+ 550,
3522
+ 555,
3523
+ 560,
3524
+ 565,
3525
+ 570,
3526
+ 575,
3527
+ 580,
3528
+ 585,
3529
+ 590,
3530
+ 595,
3531
+ 600,
3532
+ 605,
3533
+ 610,
3534
+ 615,
3535
+ 620,
3536
+ 625,
3537
+ 630,
3538
+ 635,
3539
+ 640,
3540
+ 645,
3541
+ 650,
3542
+ 655,
3543
+ 660,
3544
+ 665,
3545
+ 670,
3546
+ 675,
3547
+ 680,
3548
+ 685,
3549
+ 690,
3550
+ 695,
3551
+ 700,
3552
+ 705,
3553
+ 710,
3554
+ 715,
3555
+ 720,
3556
+ 725,
3557
+ 730,
3558
+ 735,
3559
+ 740,
3560
+ 745,
3561
+ 750,
3562
+ 755,
3563
+ 760,
3564
+ 765,
3565
+ 770,
3566
+ 775,
3567
+ 780,
3568
+ 785,
3569
+ 790,
3570
+ 795,
3571
+ 800,
3572
+ 805,
3573
+ 810,
3574
+ 815,
3575
+ 820,
3576
+ 825,
3577
+ 830,
3578
+ 835,
3579
+ 840,
3580
+ 845,
3581
+ 850,
3582
+ 855,
3583
+ 860,
3584
+ 865,
3585
+ 870,
3586
+ 875,
3587
+ 880,
3588
+ 885,
3589
+ 890,
3590
+ 895,
3591
+ 900,
3592
+ 905,
3593
+ 910,
3594
+ 915,
3595
+ 920,
3596
+ 925,
3597
+ 930,
3598
+ 935,
3599
+ 940,
3600
+ 945,
3601
+ 950,
3602
+ 955,
3603
+ 960
3604
+ ],
3605
+ "decoded_range": [
3606
+ 0,
3607
+ 962
3608
+ ],
3609
+ "clusters": [
3610
+ [
3611
+ [
3612
+ 169.21488432410015,
3613
+ 117.69717190579932
3614
+ ],
3615
+ 105
3616
+ ],
3617
+ [
3618
+ [
3619
+ 102.61586257309942,
3620
+ 132.1027046783626
3621
+ ],
3622
+ 19
3623
+ ],
3624
+ [
3625
+ [
3626
+ 89.0272393302096,
3627
+ 117.83149126439284
3628
+ ],
3629
+ 5
3630
+ ]
3631
+ ],
3632
+ "selected": [],
3633
+ "inferred_grid": [
3634
+ [
3635
+ 169.21488432410015,
3636
+ 117.69717190579932
3637
+ ],
3638
+ [
3639
+ 102.61586257309942,
3640
+ 132.1027046783626
3641
+ ],
3642
+ [
3643
+ 117.02139534566271,
3644
+ 198.70172642936336
3645
+ ],
3646
+ [
3647
+ 183.62041709666346,
3648
+ 184.29619365680006
3649
+ ]
3650
+ ],
3651
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
3652
+ "observed_side_length": 68.13918896385299,
3653
+ "other_cluster_residuals": [
3654
+ 19.70579138081687
3655
+ ]
3656
+ },
3657
+ {
3658
+ "demo": "demo_19",
3659
+ "processed_frames": [
3660
+ 25,
3661
+ 30,
3662
+ 35,
3663
+ 40,
3664
+ 45,
3665
+ 50,
3666
+ 55,
3667
+ 60,
3668
+ 65,
3669
+ 70,
3670
+ 75,
3671
+ 80,
3672
+ 85,
3673
+ 90,
3674
+ 95,
3675
+ 100,
3676
+ 105,
3677
+ 110,
3678
+ 115,
3679
+ 120,
3680
+ 125,
3681
+ 130,
3682
+ 135,
3683
+ 140,
3684
+ 145,
3685
+ 150,
3686
+ 155,
3687
+ 160,
3688
+ 165,
3689
+ 170,
3690
+ 175,
3691
+ 180,
3692
+ 185,
3693
+ 190,
3694
+ 195,
3695
+ 200,
3696
+ 205,
3697
+ 210,
3698
+ 215,
3699
+ 220,
3700
+ 225,
3701
+ 230,
3702
+ 235,
3703
+ 240,
3704
+ 245,
3705
+ 250,
3706
+ 255,
3707
+ 260,
3708
+ 265,
3709
+ 270,
3710
+ 275,
3711
+ 280,
3712
+ 285,
3713
+ 290,
3714
+ 295,
3715
+ 300,
3716
+ 305,
3717
+ 310,
3718
+ 315,
3719
+ 320,
3720
+ 325,
3721
+ 330,
3722
+ 335,
3723
+ 340,
3724
+ 345,
3725
+ 350,
3726
+ 355,
3727
+ 360,
3728
+ 365,
3729
+ 370,
3730
+ 375,
3731
+ 380,
3732
+ 385,
3733
+ 390,
3734
+ 395,
3735
+ 400,
3736
+ 405,
3737
+ 410,
3738
+ 415,
3739
+ 420,
3740
+ 425,
3741
+ 430,
3742
+ 435,
3743
+ 440,
3744
+ 445,
3745
+ 450,
3746
+ 455,
3747
+ 460,
3748
+ 465,
3749
+ 470,
3750
+ 475,
3751
+ 480,
3752
+ 485,
3753
+ 490,
3754
+ 495,
3755
+ 500,
3756
+ 505,
3757
+ 510,
3758
+ 515,
3759
+ 520,
3760
+ 525,
3761
+ 530,
3762
+ 535,
3763
+ 540,
3764
+ 545,
3765
+ 550,
3766
+ 555,
3767
+ 560,
3768
+ 565,
3769
+ 570,
3770
+ 575,
3771
+ 580,
3772
+ 585,
3773
+ 590,
3774
+ 595,
3775
+ 600,
3776
+ 605,
3777
+ 610,
3778
+ 615,
3779
+ 620,
3780
+ 625,
3781
+ 630,
3782
+ 635,
3783
+ 640,
3784
+ 645,
3785
+ 650,
3786
+ 655,
3787
+ 660,
3788
+ 665,
3789
+ 670,
3790
+ 675,
3791
+ 680,
3792
+ 685,
3793
+ 690,
3794
+ 695,
3795
+ 700,
3796
+ 705,
3797
+ 710,
3798
+ 715,
3799
+ 720,
3800
+ 725,
3801
+ 730,
3802
+ 735,
3803
+ 740,
3804
+ 745,
3805
+ 750,
3806
+ 755,
3807
+ 760,
3808
+ 765,
3809
+ 770,
3810
+ 775,
3811
+ 780,
3812
+ 785,
3813
+ 790,
3814
+ 795,
3815
+ 800,
3816
+ 805,
3817
+ 810,
3818
+ 815,
3819
+ 820,
3820
+ 825,
3821
+ 830,
3822
+ 835,
3823
+ 840,
3824
+ 845,
3825
+ 850,
3826
+ 855,
3827
+ 860,
3828
+ 865,
3829
+ 870,
3830
+ 875,
3831
+ 880,
3832
+ 885,
3833
+ 890,
3834
+ 895,
3835
+ 900,
3836
+ 905,
3837
+ 910,
3838
+ 915,
3839
+ 920,
3840
+ 925,
3841
+ 930,
3842
+ 935,
3843
+ 940,
3844
+ 945,
3845
+ 950,
3846
+ 955,
3847
+ 960,
3848
+ 965,
3849
+ 970,
3850
+ 975,
3851
+ 980,
3852
+ 985,
3853
+ 990,
3854
+ 995
3855
+ ],
3856
+ "decoded_range": [
3857
+ 0,
3858
+ 999
3859
+ ],
3860
+ "clusters": [
3861
+ [
3862
+ [
3863
+ 168.54488551126894,
3864
+ 200.20069588547275
3865
+ ],
3866
+ 93
3867
+ ],
3868
+ [
3869
+ [
3870
+ 102.66133557800234,
3871
+ 182.04100610240943
3872
+ ],
3873
+ 66
3874
+ ],
3875
+ [
3876
+ [
3877
+ 107.29696488696489,
3878
+ 97.78480278980282
3879
+ ],
3880
+ 15
3881
+ ],
3882
+ [
3883
+ [
3884
+ 100.01525715999401,
3885
+ 99.79482532845105
3886
+ ],
3887
+ 9
3888
+ ],
3889
+ [
3890
+ [
3891
+ 101.39331841432225,
3892
+ 103.54458653026427
3893
+ ],
3894
+ 4
3895
+ ]
3896
+ ],
3897
+ "selected": [],
3898
+ "inferred_grid": [
3899
+ [
3900
+ 168.54488551126894,
3901
+ 200.20069588547275
3902
+ ],
3903
+ [
3904
+ 102.66133557800234,
3905
+ 182.04100610240943
3906
+ ],
3907
+ [
3908
+ 120.82102536106567,
3909
+ 116.15745616914282
3910
+ ],
3911
+ [
3912
+ 186.70457529433227,
3913
+ 134.31714595220615
3914
+ ]
3915
+ ],
3916
+ "grid_method": "square from two most persistent adjacent cap detections; side chosen by initial cloth centroid",
3917
+ "observed_side_length": 68.3404454538184,
3918
+ "other_cluster_residuals": [
3919
+ 22.813474174366963,
3920
+ 26.469145782693513,
3921
+ 23.162907367942534
3922
+ ]
3923
+ }
3924
+ ]
agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/sequences_provenance.json ADDED
@@ -0,0 +1,122 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "demo": "demo_00",
4
+ "numeric": "all authorized action/qpos/qvel rows",
5
+ "image_frame": 25,
6
+ "camera": "top_camera"
7
+ },
8
+ {
9
+ "demo": "demo_01",
10
+ "numeric": "all authorized action/qpos/qvel rows",
11
+ "image_frame": 25,
12
+ "camera": "top_camera"
13
+ },
14
+ {
15
+ "demo": "demo_02",
16
+ "numeric": "all authorized action/qpos/qvel rows",
17
+ "image_frame": 25,
18
+ "camera": "top_camera"
19
+ },
20
+ {
21
+ "demo": "demo_03",
22
+ "numeric": "all authorized action/qpos/qvel rows",
23
+ "image_frame": 25,
24
+ "camera": "top_camera"
25
+ },
26
+ {
27
+ "demo": "demo_04",
28
+ "numeric": "all authorized action/qpos/qvel rows",
29
+ "image_frame": 25,
30
+ "camera": "top_camera"
31
+ },
32
+ {
33
+ "demo": "demo_05",
34
+ "numeric": "all authorized action/qpos/qvel rows",
35
+ "image_frame": 25,
36
+ "camera": "top_camera"
37
+ },
38
+ {
39
+ "demo": "demo_06",
40
+ "numeric": "all authorized action/qpos/qvel rows",
41
+ "image_frame": 25,
42
+ "camera": "top_camera"
43
+ },
44
+ {
45
+ "demo": "demo_07",
46
+ "numeric": "all authorized action/qpos/qvel rows",
47
+ "image_frame": 25,
48
+ "camera": "top_camera"
49
+ },
50
+ {
51
+ "demo": "demo_08",
52
+ "numeric": "all authorized action/qpos/qvel rows",
53
+ "image_frame": 25,
54
+ "camera": "top_camera"
55
+ },
56
+ {
57
+ "demo": "demo_09",
58
+ "numeric": "all authorized action/qpos/qvel rows",
59
+ "image_frame": 25,
60
+ "camera": "top_camera"
61
+ },
62
+ {
63
+ "demo": "demo_10",
64
+ "numeric": "all authorized action/qpos/qvel rows",
65
+ "image_frame": 25,
66
+ "camera": "top_camera"
67
+ },
68
+ {
69
+ "demo": "demo_11",
70
+ "numeric": "all authorized action/qpos/qvel rows",
71
+ "image_frame": 25,
72
+ "camera": "top_camera"
73
+ },
74
+ {
75
+ "demo": "demo_12",
76
+ "numeric": "all authorized action/qpos/qvel rows",
77
+ "image_frame": 25,
78
+ "camera": "top_camera"
79
+ },
80
+ {
81
+ "demo": "demo_13",
82
+ "numeric": "all authorized action/qpos/qvel rows",
83
+ "image_frame": 25,
84
+ "camera": "top_camera"
85
+ },
86
+ {
87
+ "demo": "demo_14",
88
+ "numeric": "all authorized action/qpos/qvel rows",
89
+ "image_frame": 25,
90
+ "camera": "top_camera"
91
+ },
92
+ {
93
+ "demo": "demo_15",
94
+ "numeric": "all authorized action/qpos/qvel rows",
95
+ "image_frame": 25,
96
+ "camera": "top_camera"
97
+ },
98
+ {
99
+ "demo": "demo_16",
100
+ "numeric": "all authorized action/qpos/qvel rows",
101
+ "image_frame": 25,
102
+ "camera": "top_camera"
103
+ },
104
+ {
105
+ "demo": "demo_17",
106
+ "numeric": "all authorized action/qpos/qvel rows",
107
+ "image_frame": 25,
108
+ "camera": "top_camera"
109
+ },
110
+ {
111
+ "demo": "demo_18",
112
+ "numeric": "all authorized action/qpos/qvel rows",
113
+ "image_frame": 25,
114
+ "camera": "top_camera"
115
+ },
116
+ {
117
+ "demo": "demo_19",
118
+ "numeric": "all authorized action/qpos/qvel rows",
119
+ "image_frame": 25,
120
+ "camera": "top_camera"
121
+ }
122
+ ]
agent_d20_e20/frozen_20261102/tasks/BuildTent-v1/frozen/worker_api.py ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Standalone worker client; safe to mount without the runner package."""
2
+ import base64
3
+ import json
4
+ import socket
5
+ import struct
6
+ import uuid
7
+ import os
8
+ from pathlib import Path
9
+
10
+ import numpy as np
11
+
12
+
13
+ def array_hook(value):
14
+ if value.get('__array__') is True:
15
+ dtype=np.dtype(value['dtype'])
16
+ if dtype.kind not in 'bifu':raise ValueError('non-numeric array from service')
17
+ raw=base64.b64decode(value['data'],validate=True)
18
+ return np.frombuffer(raw,dtype=dtype).reshape(value['shape']).copy()
19
+ return value
20
+
21
+
22
+ class API:
23
+ def __init__(self,socket_path='/rpc/control.sock',demo_root='/demonstrations',manifest='/demo_manifest.json'):
24
+ self.socket_path=socket_path
25
+ self.demo_root=Path(demo_root)
26
+ self.manifest=Path(manifest)
27
+
28
+ def call(self,method,arguments=None,request_id=None):
29
+ # Caller can persist a request ID before sending. Never automatically retry.
30
+ request_id=request_id or str(uuid.uuid4())
31
+ request={'request_id':request_id,'method':method,'arguments':arguments or {}}
32
+ if os.environ.get('AGENT_E20_RPC_TOKEN'):
33
+ request['token']=os.environ['AGENT_E20_RPC_TOKEN']
34
+ payload=json.dumps(request,allow_nan=False).encode()
35
+ if len(payload)>65536:raise ValueError('request too large')
36
+ with socket.socket(socket.AF_UNIX,socket.SOCK_STREAM) as s:
37
+ s.settimeout(1810);s.connect(self.socket_path)
38
+ s.sendall(struct.pack('!I',len(payload))+payload)
39
+ def exact(n):
40
+ chunks=[]
41
+ while n:
42
+ chunk=s.recv(min(n,1024*1024))
43
+ if not chunk:raise RuntimeError('RPC interrupted; do not replay action with a new ID')
44
+ chunks.append(chunk);n-=len(chunk)
45
+ return b''.join(chunks)
46
+ length=struct.unpack('!I',exact(4))[0]
47
+ if length>64*1024*1024:raise RuntimeError('response too large')
48
+ result=json.loads(exact(length),object_hook=array_hook)
49
+ if not result['ok']:raise RuntimeError(result['error'])
50
+ return result['result']
51
+
52
+ def next_episode(self,request_id=None):return self.call('next_episode',request_id=request_id)
53
+ def latest_observation(self):return self.call('latest_observation')
54
+ def execute_actions(self,actions,request_id=None):
55
+ return self.call('execute_actions',{'actions':np.asarray(actions).tolist()},request_id)
56
+ def finish_episode(self):return self.call('finish_episode')
57
+ def budget_status(self):return self.call('budget_status')
58
+ def rehearse(self,candidate='candidate',request_id=None):
59
+ return self.call('rehearse',{'candidate':candidate},request_id)
60
+ def score_rehearsal(self,episode,visible_success,evidence):
61
+ return self.call('score_rehearsal',{'episode':episode,'visible_success':visible_success,'evidence':evidence})
62
+ def freeze(self):return self.call('freeze')
63
+
64
+ def read_demo(self,demo_id,start=0,stop=None,fields=None):
65
+ import h5py
66
+ manifest=json.loads(self.manifest.read_text())
67
+ if manifest.get('status') not in ('released','released_for_offline_analysis'):
68
+ raise RuntimeError('demonstrations not released by trusted preflight')
69
+ entries={e['demo_id']:e for e in manifest['episodes']}
70
+ if demo_id not in entries:raise ValueError('unauthorized demo ID')
71
+ entry=entries[demo_id]
72
+ allowed={f['path'] for f in entry['fields']}
73
+ fields=sorted(allowed) if fields is None else fields
74
+ if not set(fields)<=allowed:raise ValueError('unauthorized demo field')
75
+ stop=entry['length']+1 if stop is None else stop
76
+ if not isinstance(start,int) or not isinstance(stop,int) or not 0<=start<stop<=entry['length']+1:
77
+ raise ValueError('invalid frame interval')
78
+ path=(self.demo_root/entry['file']).resolve()
79
+ if path.parent!=self.demo_root.resolve():raise ValueError('invalid demo file')
80
+ with h5py.File(path,'r') as f:
81
+ return {field:f[field][start:min(stop,len(f[field]))] for field in fields}
agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/__pycache__/build_bank.cpython-310.pyc ADDED
Binary file (2.29 kB). View file
 
agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/__pycache__/check_synthetic.cpython-310.pyc ADDED
Binary file (2.91 kB). View file
 
agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/__pycache__/controller.cpython-310.pyc ADDED
Binary file (2.25 kB). View file
 
agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/__pycache__/features.cpython-310.pyc ADDED
Binary file (1.61 kB). View file
 
agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/controller.py ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """D20 closed-loop retrieval with phase advancement: observation-conditioned local action retrieval."""
2
+ from pathlib import Path
3
+ import importlib.util
4
+ import numpy as np
5
+ _spec=importlib.util.spec_from_file_location('d20_features',Path(__file__).with_name('features.py'))
6
+ _f=importlib.util.module_from_spec(_spec);_spec.loader.exec_module(_f)
7
+ _ss=importlib.util.spec_from_file_location('d20_servo',Path(__file__).with_name('servo.py'))
8
+ _servo=importlib.util.module_from_spec(_ss);_ss.loader.exec_module(_servo)
9
+ _bank=None
10
+ _progress=0.
11
+ _terminal_count=0
12
+ _last_step=0
13
+ _finish=0
14
+
15
+ def reset_episode():
16
+ global _progress,_terminal_count,_last_step,_finish
17
+ _progress=0.;_terminal_count=0;_last_step=0;_finish=0
18
+
19
+ def act(obs):
20
+ global _bank,_progress,_terminal_count,_last_step,_finish
21
+ step=int(obs['step_id'])
22
+ # Step 1 follows the runner-owned hold; at most 299 further native actions.
23
+ if step>=300:return None
24
+ if step<1:raise ValueError('step_id must start at 1 after runner hold')
25
+ if step<=_last_step:reset_episode()
26
+ _last_step=step
27
+ if _progress>=2.75 or _finish:
28
+ _finish+=1
29
+ if _finish>45:return None
30
+ a=np.zeros(14,np.float32);a[[6,13]]=1
31
+ if 8<_finish<=28:a[[2,9]]=.08
32
+ return a
33
+ if _bank is None:
34
+ with np.load(Path(__file__).with_name('reference_bank.npz'),allow_pickle=False) as z:
35
+ _bank={k:z[k].astype(np.float32) for k in z.files}
36
+ rates=np.r_[np.diff(_bank['phase']),.018]
37
+ rates[rates<=0]=.018
38
+ _bank['rate']=rates
39
+ try:q,v,area=_f.features(obs)
40
+ except (KeyError,ValueError,TypeError):return None
41
+ b=_bank
42
+ # Proprioception locates the arms; cloth masks and RGB distinguish scene geometry.
43
+ dq=np.mean((b['q']-q)**2,axis=1)
44
+ dv=np.mean((b['v']-v)**2,axis=1)
45
+ cost=dq+4*dv
46
+ # Weak continuity regularizer; all demos remain eligible on every observation.
47
+ cost+=.8*(b['phase']-_progress)**2
48
+ cost[b['phase'] < _progress-.08] += 100
49
+ order=np.argsort(cost)[:8];best=int(order[0])
50
+ if dq[best]>9 or dv[best]>.8:return None # unsupported observation, not success
51
+ # Avoid averaging contradictory gripper modes at an event boundary.
52
+ mode=b['action'][best,[6,13]]
53
+ idx=order[np.all(b['action'][order][:,[6,13]]==mode,axis=1)]
54
+ w=np.exp(-(cost[idx]-cost[best])/.035);w/=w.sum()
55
+ action=np.sum(b['action'][idx]*w[:,None],axis=0)
56
+ target=np.sum(b['q'][idx]*w[:,None],axis=0)
57
+ action+=_servo.correction(q,target)
58
+ observed=float(np.sum(b['phase'][idx]*w))
59
+ rate=float(np.sum(b['rate'][idx]*w))
60
+ _progress=float(min(3.,max(_progress+rate,min(observed,_progress+.035))))
61
+ terminal=bool(b['terminal'][best]) and dq[best]<.03 and dv[best]<.025
62
+ _terminal_count=_terminal_count+1 if terminal else 0
63
+ if _terminal_count>=5:return None # observation resembles demo endpoint; unverified completion
64
+ action=np.clip(action,-1,1).astype(np.float32)
65
+ if action.shape!=(14,) or not np.isfinite(action).all():return None
66
+ return action
agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/environment.lock ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ absl-py==2.4.0
2
+ accelerate==1.13.0
3
+ addict==2.4.0
4
+ aiohappyeyeballs==2.6.1
5
+ aiohttp==3.13.5
6
+ aiosignal==1.4.0
7
+ annotated-types==0.7.0
8
+ anyio==4.13.0
9
+ anytree==2.13.0
10
+ asttokens==3.0.1
11
+ async-timeout==5.0.1
12
+ attrs==26.1.0
13
+ av==15.1.0
14
+ blinker==1.9.0
15
+ casadi==3.7.2
16
+ certifi==2026.4.22
17
+ cffi==2.0.0
18
+ charset-normalizer==3.4.7
19
+ click==8.3.3
20
+ cloudpickle==3.1.2
21
+ cmake==4.1.3
22
+ coacd==1.0.11
23
+ colorama==0.4.6
24
+ colorlog==6.10.1
25
+ comm==0.2.3
26
+ ConfigArgParse==1.7.5
27
+ contourpy==1.3.2
28
+ cuda-bindings==12.9.4
29
+ cuda-pathfinder==1.5.4
30
+ cycler==0.12.1
31
+ dacite==1.9.2
32
+ dash==4.1.0
33
+ dataclasses-json==0.6.7
34
+ datasets==4.8.5
35
+ decorator==5.2.1
36
+ deepdiff==8.6.2
37
+ Deprecated==1.3.1
38
+ diffusers==0.35.2
39
+ dill==0.4.1
40
+ docstring_parser==0.18.0
41
+ draccus==0.10.0
42
+ einops==0.8.2
43
+ embreex==4.4.0
44
+ etils==1.13.0
45
+ evdev==1.9.3
46
+ exceptiongroup==1.3.1
47
+ executing==2.2.1
48
+ Farama-Notifications==0.0.6
49
+ fast_simplification==0.1.13
50
+ fastjsonschema==2.21.2
51
+ filelock==3.29.0
52
+ Flask==3.1.3
53
+ fonttools==4.63.0
54
+ freetype-py==2.5.1
55
+ frozenlist==1.8.0
56
+ fsspec==2026.2.0
57
+ genesis-world @ file:///data/gck/genesis-world
58
+ git-filter-repo==2.47.0
59
+ gitdb==4.0.12
60
+ GitPython==3.1.50
61
+ glfw==2.10.0
62
+ gs-madrona==0.0.5
63
+ gstaichi==3.3.0
64
+ gym==0.26.2
65
+ gym-notices==0.1.0
66
+ gymnasium==1.3.0
67
+ h11==0.16.0
68
+ h5py==3.16.0
69
+ hf-xet==1.5.0
70
+ hf_transfer==0.1.9
71
+ httpcore==1.0.9
72
+ httpx==0.28.1
73
+ huggingface-hub==0.35.3
74
+ idna==3.15
75
+ ImageIO==2.37.3
76
+ imageio-ffmpeg==0.6.0
77
+ importlib_metadata==9.0.0
78
+ importlib_resources==7.1.0
79
+ inquirerpy==0.3.4
80
+ ipython==8.39.0
81
+ ipywidgets==8.1.8
82
+ itsdangerous==2.2.0
83
+ jedi==0.20.0
84
+ Jinja2==3.1.6
85
+ joblib==1.5.3
86
+ jsonlines==4.0.0
87
+ jsonschema==4.26.0
88
+ jsonschema-specifications==2025.9.1
89
+ jupyter_core==5.9.1
90
+ jupyterlab_widgets==3.0.16
91
+ kiwisolver==1.5.0
92
+ lazy-loader==0.5
93
+ lerobot==0.4.4
94
+ libigl==2.6.2
95
+ llvmlite==0.47.0
96
+ loguru==0.7.3
97
+ lxml==6.1.0
98
+ manifold3d==3.4.1
99
+ -e git+ssh://git@github.com/HeegerGao/maniladder.git@f6eed17496816bb41ce484f2d7580990a908d3e6#egg=maniladder
100
+ mapbox_earcut==2.0.0
101
+ markdown-it-py==4.2.0
102
+ MarkupSafe==3.0.3
103
+ marshmallow==3.26.2
104
+ matplotlib==3.10.9
105
+ matplotlib-inline==0.2.2
106
+ mdurl==0.1.2
107
+ mergedeep==1.3.4
108
+ moviepy==2.2.1
109
+ mpmath==1.3.0
110
+ msgspec==0.21.1
111
+ mujoco==3.8.1
112
+ multidict==6.7.1
113
+ multiprocess==0.70.19
114
+ mypy_extensions==1.1.0
115
+ nbformat==5.10.4
116
+ nest-asyncio==1.6.0
117
+ networkx==3.4.2
118
+ ninja==1.13.0
119
+ nlopt==2.10.0
120
+ numba==0.65.1
121
+ numpy @ file:///home/conda/feedstock_root/build_artifacts/numpy_1747544640217/work/dist/numpy-2.2.6-cp310-cp310-linux_x86_64.whl#sha256=d6d964caeef85d00073d27cd62b46883d275b3d8162f723f0fcabbd0b3cc3f9d
122
+ nvidia-cublas-cu12==12.8.4.1
123
+ nvidia-cuda-cupti-cu12==12.8.90
124
+ nvidia-cuda-nvrtc-cu12==12.8.93
125
+ nvidia-cuda-runtime-cu12==12.8.90
126
+ nvidia-cudnn-cu12==9.10.2.21
127
+ nvidia-cufft-cu12==11.3.3.83
128
+ nvidia-cufile-cu12==1.13.1.3
129
+ nvidia-curand-cu12==10.3.9.90
130
+ nvidia-cusolver-cu12==11.7.3.90
131
+ nvidia-cusparse-cu12==12.5.8.93
132
+ nvidia-cusparselt-cu12==0.7.1
133
+ nvidia-nccl-cu12==2.27.5
134
+ nvidia-nvjitlink-cu12==12.8.93
135
+ nvidia-nvshmem-cu12==3.4.5
136
+ nvidia-nvtx-cu12==12.8.90
137
+ open3d==0.19.0
138
+ opencv-python==4.13.0.92
139
+ opencv-python-headless==4.12.0.88
140
+ OpenEXR==3.4.11
141
+ orderly-set==5.5.0
142
+ packaging==25.0
143
+ pandas==2.3.3
144
+ parso==0.8.7
145
+ pexpect==4.9.0
146
+ pfzy==0.3.4
147
+ pillow==11.3.0
148
+ pin==4.0.0
149
+ pip @ file:///home/task_177498883471825/croot/pip_1774988882575/work
150
+ platformdirs==4.9.6
151
+ plotly==5.24.1
152
+ polyscope==2.6.1
153
+ proglog==0.1.12
154
+ prompt_toolkit==3.0.52
155
+ propcache==0.5.2
156
+ protobuf==6.33.6
157
+ psutil==7.2.2
158
+ ptyprocess==0.7.0
159
+ pure_eval==0.2.3
160
+ py-cpuinfo==9.0.0
161
+ py-spy==0.4.2
162
+ pyarrow==24.0.0
163
+ pycollada==0.9.3
164
+ pycparser==3.0
165
+ pydantic==2.13.4
166
+ pydantic_core==2.46.4
167
+ PyGEL3D==0.6.1
168
+ pyglet==2.1.14
169
+ pygltflib==1.16.0
170
+ Pygments==2.20.0
171
+ pymeshlab==2025.7.post1
172
+ pynput==1.8.2
173
+ PyOpenGL==3.1.10
174
+ pyparsing==3.3.2
175
+ pyperclip==1.11.0
176
+ PyQt6==6.11.0
177
+ PyQt6-Qt6==6.11.0
178
+ PyQt6_sip==13.11.1
179
+ pyquaternion==0.9.9
180
+ pyserial==3.5
181
+ pysplashsurf==0.12.0.2
182
+ python-dateutil==2.9.0.post0
183
+ python-dotenv==1.2.2
184
+ python-xlib==0.33
185
+ pytransform3d==3.15.0
186
+ pytz==2026.2
187
+ pyuipc @ file:///tmp/claude-1005/-mnt4-gck-maniladder/9f005e67-04f6-4368-bd0e-9f40896c7dbf/scratchpad/pyuipc_dl/pyuipc-0.0.25-cp310-cp310-manylinux_2_34_x86_64.whl#sha256=76ff13554e1971ea4401db2c875369cf32352ae3a5a4c398c45193231b5cbf00
188
+ PyYAML==6.0.3
189
+ pyyaml-include==1.4.1
190
+ referencing==0.37.0
191
+ regex==2026.5.9
192
+ requests==2.34.2
193
+ rerun-sdk==0.26.2
194
+ retrying==1.4.2
195
+ rich==14.3.4
196
+ rpds-py==0.30.0
197
+ rtree==1.4.1
198
+ safetensors==0.7.0
199
+ sapien==3.0.3
200
+ scikit-image==0.25.2
201
+ scikit-learn==1.7.2
202
+ scipy @ file:///home/conda/feedstock_root/build_artifacts/scipy-split_1739790642651/work/dist/scipy-1.15.2-cp310-cp310-linux_x86_64.whl#sha256=9e52bad6c3294d1a5b04a13632118ca2157130603c6c018c2d710162b223b27e
203
+ sentry-sdk==2.60.0
204
+ setuptools==80.10.2
205
+ shapely==2.1.2
206
+ six==1.17.0
207
+ smmap==5.0.3
208
+ stack-data==0.6.3
209
+ svg.path==7.0
210
+ sympy==1.14.0
211
+ tenacity==9.1.4
212
+ termcolor==3.3.0
213
+ tetgen==0.8.4
214
+ threadpoolctl==3.6.0
215
+ tifffile==2025.5.10
216
+ timm==1.0.27
217
+ toml==0.10.2
218
+ torch==2.10.0
219
+ torchaudio==2.11.0
220
+ torchcodec==0.10.0
221
+ torchvision==0.25.0
222
+ tqdm==4.67.3
223
+ traitlets==5.15.0
224
+ transforms3d==0.4.2
225
+ trimesh==4.12.2
226
+ triton==3.6.0
227
+ typeguard==4.5.2
228
+ typing-inspect==0.9.0
229
+ typing-inspection==0.4.2
230
+ typing_extensions==4.15.0
231
+ tyro==1.0.13
232
+ tzdata==2026.2
233
+ urllib3==2.7.0
234
+ vhacdx==0.0.10
235
+ viser==1.0.27
236
+ vtk==9.6.1
237
+ wandb==0.24.2
238
+ wcwidth==0.7.0
239
+ websockets==16.0
240
+ Werkzeug==3.1.8
241
+ wheel==0.46.3
242
+ widgetsnbextension==4.0.15
243
+ wrapt==2.1.2
244
+ xxhash==3.7.0
245
+ yarl==1.23.0
246
+ yourdfpy==0.0.60
247
+ z3-solver==4.16.0.0
248
+ zipp==3.23.1
249
+ zstandard==0.25.0
agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/features.py ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Fixed, non-neural observation features. No fitted or pretrained weights."""
2
+ import cv2
3
+ import numpy as np
4
+ CAMERAS=('top_camera','front_camera')
5
+ PATHS=tuple(f'agent/panda-{r}/{k}' for r in range(2) for k in ('qpos','qvel'))
6
+ def features(obs):
7
+ q=[]
8
+ for p in PATHS:
9
+ a=np.asarray(obs['proprioception'][p],dtype=np.float32).reshape(-1)
10
+ if a.size!=9 or not np.isfinite(a).all(): raise ValueError('invalid proprioception: '+p)
11
+ scale=np.array([.35]*7+[.025]*2) if p.endswith('qpos') else np.array([2.]*9)
12
+ q.extend(a/scale)
13
+ visual=[]; area=[]
14
+ for c in CAMERAS:
15
+ a=np.asarray(obs['rgb'][c])
16
+ if a.ndim==4 and a.shape[0]==1:a=a[0]
17
+ if a.ndim!=3 or a.shape[2]!=3 or a.dtype!=np.uint8:raise ValueError('invalid RGB: '+c)
18
+ rgb=a.astype(np.float32)/255
19
+ mask=((rgb[:,:,0]>.35)&(rgb[:,:,1]>.30)&(rgb[:,:,2]<.55*np.minimum(rgb[:,:,0],rgb[:,:,1]))).astype(np.float32)
20
+ area.append(float(mask.mean()))
21
+ visual.extend(cv2.resize(mask,(24,24),interpolation=cv2.INTER_AREA).reshape(-1)*3)
22
+ visual.extend(cv2.resize(rgb,(12,12),interpolation=cv2.INTER_AREA).reshape(-1))
23
+ return np.asarray(q,np.float32),np.asarray(visual,np.float32),np.asarray(area,np.float32)
agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/freeze_manifest.json ADDED
@@ -0,0 +1,142 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "files": {
3
+ "__pycache__/build_bank.cpython-310.pyc": "487cc933211ddc843679c283222c3cd29905d39a28b9b3ac74117f91fc053871",
4
+ "__pycache__/check_synthetic.cpython-310.pyc": "91d08b50c8e781309cb979e310502ba1274d69cf97efad7da0d16f2c215a0855",
5
+ "__pycache__/controller.cpython-310.pyc": "fe1ecff60121c49cd8f89699d5daaab398c943fa5dac98282e94441af4ad5ab2",
6
+ "__pycache__/features.cpython-310.pyc": "62faf4c65dafa1b5345ab29297241732ed73d13eec46644dcf9535e6326be21e",
7
+ "build_bank.py": "32d4ff3433cfd1ce63d7ebdc5893952128a7862d7798954d2f2d67154f65356b",
8
+ "check_synthetic.py": "44d9b772f8a2341d03f65e1ff6956f4bf2f95705e092cd0240d3551b65e069a1",
9
+ "controller.py": "8b30242d71607427c6d7174c517743f83faa98a13918257f44e886fb9da11d7e",
10
+ "demo_manifest.json": "ddd91490ba609d1c2d9bc513800145733b7e9cffbfe5de78a63a35b0bcb755bf",
11
+ "demonstrations_readonly/demo_00.h5": "86d12baa67226bbce45df91f6f4bf6a42211454bb0b2530f5ecb5798deba7923",
12
+ "demonstrations_readonly/demo_01.h5": "f6feaae38a42aef0fc87ecb572f7856cc9815e92a64b733817b479e1a5d313b3",
13
+ "demonstrations_readonly/demo_02.h5": "41442c9de6cbec0e18c328b8c887a9fa718bdb6429f5d0ed092020134bd1e950",
14
+ "demonstrations_readonly/demo_03.h5": "971d17a10c3c03368ac7ddb658c76c641ac79c8eb143d912e7fa688a2956d712",
15
+ "demonstrations_readonly/demo_04.h5": "d1c69c4ef6648d80807e7f152e56f578e4ff66204ff26246d84e2124d857a8f9",
16
+ "demonstrations_readonly/demo_05.h5": "cbd22f18a95edb6616339cc5d4a4cba2a4d8f9dc3e68ca5dda206b078bee05df",
17
+ "demonstrations_readonly/demo_06.h5": "e511e007095c2ee27d1d784c06e80d2adefb3eee51fc5d795bf8b39e8fe4e5eb",
18
+ "demonstrations_readonly/demo_07.h5": "ad31e5e2ced962e40029d678ebe9088f55cccc1c417574088f2ce0d3373db2e4",
19
+ "demonstrations_readonly/demo_08.h5": "e6bdbdc6294b5780dcb424549d85f1a36693bfa74476fcc93aa91dbe41f467e0",
20
+ "demonstrations_readonly/demo_09.h5": "7db3675ad504c2c24dd58216246d1e8476a4bbdafa51ab64c3288b73e34ea1b0",
21
+ "demonstrations_readonly/demo_10.h5": "9b45758b14889ea559055c57814a2e9b94d00c876a8edede1446a0e39f6fb4db",
22
+ "demonstrations_readonly/demo_11.h5": "14939697824e6d08d6dfd648b906885f80bf87b7378eab0af13038696e83a241",
23
+ "demonstrations_readonly/demo_12.h5": "04c92be1d6924fa9f0799187a9b07b6f215c94dceaaf3531eea94df3986ec42e",
24
+ "demonstrations_readonly/demo_13.h5": "10b787ce6f5251fea712d9931d6644ce54fc7934141844d55b836d03f3f24cae",
25
+ "demonstrations_readonly/demo_14.h5": "49ad3137e5f12e607be76b780dfe0ecba47e8ea8fa7ca579ffc9b922f177f38c",
26
+ "demonstrations_readonly/demo_15.h5": "e237032a791b021736c6dd4ed5d73f6e91f569866c98c451e91786b8126fdffb",
27
+ "demonstrations_readonly/demo_16.h5": "90332e029671cf6c6937fc2e3bfa4ac5190c240f85363da5dc4301b691ef5691",
28
+ "demonstrations_readonly/demo_17.h5": "90ca9e1b8e1ed2f092246918126c1660ee8f3f8eeaba56b2761932cbeaa5f0bd",
29
+ "demonstrations_readonly/demo_18.h5": "5e3834be653bff024b52bd0e591ede700c8be788ccfe9833414293f9f9300188",
30
+ "demonstrations_readonly/demo_19.h5": "54fea9809ebf536bbfa0006a64b1f24d0e2f1ec85772fc393ed649c1f223ac04",
31
+ "environment.lock": "f9f11f093744865020243315494d48c69e38ff9e67783e7cf5f0a1e5f22b34aa",
32
+ "features.py": "14b124498cb63306c67a1305ce059b1205c8bde445431eeec9ff805de8f7045f",
33
+ "findings.md": "a509abae15af4fb1f525da11daa32acb019a1508afa2dbe1e3452a32eed9d7d4",
34
+ "handoff.md": "79609670a86adc4dfadb7b6a73fe6d2483cffba45c13649bb56fd66726f6abdd",
35
+ "model_tool_manifest.yaml": "584f7641d7e2d6e99db9e12854827aa4fff4ccd58238b441c080986f9861bee0",
36
+ "policy_runner.py": "cd8a30fc763583fb9bf6570414f01d56aba16241c733fc0cdfdfab0f8f27270e",
37
+ "reference_bank.npz": "e172f702629b75ab09a30cbe3cdaa503370cfe7269512f4bc58b41b47161b7d8",
38
+ "servo.py": "6eba20eafb4591f9584780078784cdb9dcc83c20c397d873cc719da640d8399b",
39
+ "servo_bank.npz": "1a49af7b71e87d28e3609d70143548927807532ec7b8b94a2d62592b26023f6c",
40
+ "skills.md": "65d68443d27b15e2e146b0ffcaa25a7697ce6c6fd5e9b634c53bac7bdeb48553",
41
+ "synthetic_check.txt": "40caf90e0875b90ab93c8f3e16a90201b0c5f966ca696ed98deb111118eff0ac",
42
+ "worker_api.py": "88cca5fa5c1c9570ca2bd90e8848f67c09224fa498d94e63634ad191e913c41e"
43
+ },
44
+ "bundle_hash": "dde4f0c6ceeba7e46dc17d28eef81722dfc16feec3a1b8dbf36cc58de85757ae",
45
+ "exploration_ledger_hash": "eece05cc136e852853df0de4d610c75879305ab9113afed1e950dd694aa79ec1",
46
+ "entrypoint": "controller.reset_episode(); controller.act(obs)",
47
+ "selection": {
48
+ "rule": "exact-version success rate, valid rehearsal count, recency",
49
+ "rehearsals": [
50
+ {
51
+ "bundle_hash": "535604fc09493af34fcbf530bde4be84d42245cb27bc522a0101d264e1960f00",
52
+ "episode": 4,
53
+ "independent": true,
54
+ "score_source": "native_step_feedback",
55
+ "status": "finished",
56
+ "success": false,
57
+ "valid": true
58
+ },
59
+ {
60
+ "bundle_hash": "dde4f0c6ceeba7e46dc17d28eef81722dfc16feec3a1b8dbf36cc58de85757ae",
61
+ "episode": 8,
62
+ "independent": true,
63
+ "score_source": "native_step_feedback",
64
+ "status": "finished",
65
+ "success": true,
66
+ "valid": true
67
+ },
68
+ {
69
+ "bundle_hash": "dde4f0c6ceeba7e46dc17d28eef81722dfc16feec3a1b8dbf36cc58de85757ae",
70
+ "episode": 9,
71
+ "independent": true,
72
+ "score_source": "native_step_feedback",
73
+ "status": "finished",
74
+ "success": false,
75
+ "valid": true
76
+ },
77
+ {
78
+ "episode": 14,
79
+ "bundle_hash": "dde4f0c6ceeba7e46dc17d28eef81722dfc16feec3a1b8dbf36cc58de85757ae",
80
+ "independent": true,
81
+ "valid": true,
82
+ "status": "finished",
83
+ "success": true,
84
+ "score_source": "native_step_feedback"
85
+ },
86
+ {
87
+ "episode": 15,
88
+ "bundle_hash": "dde4f0c6ceeba7e46dc17d28eef81722dfc16feec3a1b8dbf36cc58de85757ae",
89
+ "independent": true,
90
+ "valid": true,
91
+ "status": "finished",
92
+ "success": true,
93
+ "score_source": "native_step_feedback"
94
+ },
95
+ {
96
+ "episode": 16,
97
+ "bundle_hash": "dde4f0c6ceeba7e46dc17d28eef81722dfc16feec3a1b8dbf36cc58de85757ae",
98
+ "independent": true,
99
+ "valid": true,
100
+ "status": "finished",
101
+ "success": true,
102
+ "score_source": "native_step_feedback"
103
+ },
104
+ {
105
+ "episode": 17,
106
+ "bundle_hash": "dde4f0c6ceeba7e46dc17d28eef81722dfc16feec3a1b8dbf36cc58de85757ae",
107
+ "independent": true,
108
+ "valid": true,
109
+ "status": "finished",
110
+ "success": true,
111
+ "score_source": "native_step_feedback"
112
+ },
113
+ {
114
+ "episode": 18,
115
+ "bundle_hash": "dde4f0c6ceeba7e46dc17d28eef81722dfc16feec3a1b8dbf36cc58de85757ae",
116
+ "independent": true,
117
+ "valid": true,
118
+ "status": "finished",
119
+ "success": false,
120
+ "score_source": "native_step_feedback"
121
+ },
122
+ {
123
+ "episode": 19,
124
+ "bundle_hash": "dde4f0c6ceeba7e46dc17d28eef81722dfc16feec3a1b8dbf36cc58de85757ae",
125
+ "independent": true,
126
+ "valid": true,
127
+ "status": "finished",
128
+ "success": false,
129
+ "score_source": "native_step_feedback"
130
+ },
131
+ {
132
+ "episode": 20,
133
+ "bundle_hash": "dde4f0c6ceeba7e46dc17d28eef81722dfc16feec3a1b8dbf36cc58de85757ae",
134
+ "independent": true,
135
+ "valid": true,
136
+ "status": "finished",
137
+ "success": false,
138
+ "score_source": "native_step_feedback"
139
+ }
140
+ ]
141
+ }
142
+ }
agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/model_tool_manifest.yaml ADDED
@@ -0,0 +1,69 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "status": "E20_v5_interactive_native_success_ep7",
3
+ "task_id": "DualArmFoldHandKerchief-v1",
4
+ "weights_modified": false,
5
+ "pretrained_weight_references": [],
6
+ "model_identity": "Provider-hosted coding assistant; exact checkpoint/weight hash not exposed. No model training invoked.",
7
+ "runtime_dependencies": {
8
+ "python": "3.10.20",
9
+ "numpy": "2.2.6",
10
+ "opencv": "4.12.0"
11
+ },
12
+ "offline_dependencies": {
13
+ "python": "3.10.20",
14
+ "numpy": "2.2.6",
15
+ "opencv": "4.12.0",
16
+ "h5py": "3.16.0",
17
+ "scipy": "1.15.2"
18
+ },
19
+ "tools_used": [
20
+ "functions.exec local shell/Python",
21
+ "functions.view_image local contact sheets",
22
+ "worker_api.API.read_demo (file access only)",
23
+ "bash cat/mkdir/find; rg unavailable"
24
+ ],
25
+ "data_provenance": {
26
+ "authorized_demo_ids": [
27
+ "demo_00",
28
+ "demo_01",
29
+ "demo_02",
30
+ "demo_03",
31
+ "demo_04",
32
+ "demo_05",
33
+ "demo_06",
34
+ "demo_07",
35
+ "demo_08",
36
+ "demo_09",
37
+ "demo_10",
38
+ "demo_11",
39
+ "demo_12",
40
+ "demo_13",
41
+ "demo_14",
42
+ "demo_15",
43
+ "demo_16",
44
+ "demo_17",
45
+ "demo_18",
46
+ "demo_19"
47
+ ],
48
+ "source": "/demonstrations via /demo_manifest.json",
49
+ "contract_sha256": "103d0bc73de7b4d3ebf7eda9f4a5cbcaecbfb94bb48971be7a8a3dd9bfa2dc32",
50
+ "manifest_sha256": "40f94e0f05a4f6ba2601c405604ab7d5c71b6a4679a755476f6d6a03dbe3ecb2",
51
+ "reference_bank": "All legal top/front RGB and qpos/qvel observations; expert labels; fixed feature extraction, no neural fitting",
52
+ "servo_bank": "4744 local joint-difference/action samples derived solely from this task D20 bank; numerical least squares, no neural fitting"
53
+ },
54
+ "network": "No requests made; audited proxy availability untested",
55
+ "exploration_episodes": 7,
56
+ "environment_tests": 7,
57
+ "limitations": [
58
+ "Success established only for one interactive scene",
59
+ "Grasp localization and phase transfer remain imperfect",
60
+ "Native score is authoritative; images may be occluded"
61
+ ],
62
+ "artifact_sha256": {
63
+ "controller.py": "8b30242d71607427c6d7174c517743f83faa98a13918257f44e886fb9da11d7e",
64
+ "features.py": "14b124498cb63306c67a1305ce059b1205c8bde445431eeec9ff805de8f7045f",
65
+ "reference_bank.npz": "e172f702629b75ab09a30cbe3cdaa503370cfe7269512f4bc58b41b47161b7d8",
66
+ "servo.py": "6eba20eafb4591f9584780078784cdb9dcc83c20c397d873cc719da640d8399b",
67
+ "servo_bank.npz": "1a49af7b71e87d28e3609d70143548927807532ec7b8b94a2d62592b26023f6c"
68
+ }
69
+ }
agent_d20_e20/frozen_20261102/tasks/DualArmFoldHandKerchief-v1/frozen/skills.md ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ # Observation-feedback folding skills
2
+ Localize using yellow masks and coarse RGB from top/front cameras, with both robots' qpos/qvel. Retrieve current-observation action examples across this task's 20 authorized demonstrations, using a monotonic phase prior. Advance phase using retrieved demonstration increments to avoid waiting-pose deadlocks while respecting varying approach/fold durations.
3
+ Apply a bounded numerical pose correction from local legal-demonstration joint differences and translation actions. Each correction component is capped at .035 normalized (3.5 mm). No neural fitting, external models or networking.
4
+ Approach, close both grippers, lift/fold, and release/reposition are represented by demonstration labels. A late phase2.75 finish opens both grippers, raises hands, and settles for inspection. Finger gaps cannot certify grasp. Unsupported observations or horizon stop the program. There is no program-level success claim. One interactive native success occurred in episode7; independent evidence is recorded separately.
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/__pycache__/build_memory.cpython-310.pyc ADDED
Binary file (3.29 kB). View file
 
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/__pycache__/check_synthetic.cpython-310.pyc ADDED
Binary file (2.17 kB). View file
 
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/__pycache__/controller.cpython-310.pyc ADDED
Binary file (2.52 kB). View file
 
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/__pycache__/features.cpython-310.pyc ADDED
Binary file (1.75 kB). View file
 
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/build_memory.py ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Offline compilation from only authorized D20 observations and labels."""
2
+ import importlib.util,json
3
+ from pathlib import Path
4
+ import numpy as np
5
+ from features import visual,proprio,PATHS,CAMERAS
6
+ s=importlib.util.spec_from_file_location('worker_api','/worker_api.py');m=importlib.util.module_from_spec(s);s.loader.exec_module(m);api=m.API()
7
+ c=json.load(open('/task_contract.json'));manifest=json.load(open('/demo_manifest.json'));parts=c['action_parts']
8
+ qs=[];vs=[];actions=[];ids=[];times=[];summary=[]
9
+ for di,e in enumerate(manifest['episodes']):
10
+ n=e['length'];fields=['obs/'+p for p in PATHS]+['actions/'+p['robot_key'] for p in parts]+['obs/sensor_data/'+cam+'/rgb' for cam in CAMERAS]
11
+ d=api.read_demo(e['demo_id'],fields=fields)
12
+ q=np.concatenate([d['obs/'+p][:-1] for p in PATHS],1)
13
+ v=np.stack([visual({cam:d['obs/sensor_data/'+cam+'/rgb'][t] for cam in CAMERAS}) for t in range(n)])
14
+ a=np.concatenate([d['actions/'+p['robot_key']] for p in parts],1)
15
+ qs.append(q);vs.append(v);actions.append(a);ids.extend([di]*n);times.extend(range(n))
16
+ # All blocks retained for actual sequence analysis, no evaluation rollout.
17
+ summary.append({'id':e['demo_id'],'blocks':[{'start':t,'stop_exclusive':min(t+50,n),'arm_mean':np.mean(a[t:t+50][:,[0,1,2,3,22,23,24,25]],0).tolist(),'hand_abs_mean':float(abs(a[t:t+50][:,list(range(6,22))+list(range(28,44))]).mean())} for t in range(0,n,50)]})
18
+ print(e['demo_id'],n,flush=True)
19
+ q=np.concatenate(qs);v=np.concatenate(vs);a=np.concatenate(actions)
20
+ # Empirical scales are descriptive normalization, not neural training.
21
+ scale=np.maximum(q.std(0),np.tile(np.r_[np.full(23,.12),np.full(23,.5)],2))
22
+ np.savez_compressed(Path(__file__).with_name('memory.npz'),q=q,v=v,a=a,scale=scale,demo=np.array(ids),frame=np.array(times))
23
+ json.dump(summary,open('/workspace/evidence/sequence_blocks.json','w'),indent=2)
24
+ json.dump({'action_parts':parts,'action_low':c['action_low'],'action_high':c['action_high'],'native_horizon':c['native_horizon'],'control_hz':c['control_hz'],'demo_ids':[e['demo_id'] for e in manifest['episodes']]},open(Path(__file__).with_name('config.json'),'w'),indent=2)
25
+ print('samples',len(a),'initial q spread',np.ptp(np.stack([x[0] for x in qs]),0).max())
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/check_synthetic.py ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Interface and synthetic dependency checks only; no environment or demo rollout."""
2
+ import importlib.util
3
+ from pathlib import Path
4
+ import numpy as np
5
+ p=Path(__file__).parent
6
+ s=importlib.util.spec_from_file_location('candidate_controller',p/'controller.py');c=importlib.util.module_from_spec(s);s.loader.exec_module(c)
7
+ def obs():
8
+ return {'rgb':{'front_camera':np.zeros((240,320,3),np.uint8),'top_camera':np.zeros((1,320,240,3),np.uint8)},'proprioception':{k:np.zeros(23,np.float32) for k in c._features.PATHS},'instruction':'synthetic check','step_id':1}
9
+ o=obs();q=c._features.proprio(o['proprioception']);v=c._features.visual(o['rgb']);assert v.shape==(1184,)
10
+ # Controlled synthetic memory demonstrates that the selector uses both modalities.
11
+ q1=q.copy();q1[0]=3
12
+ v1=v.copy();v1[:]=.5
13
+ ci=np.r_[np.arange(16),np.arange(592,608)]
14
+ c._memory={'q':np.stack([q,q1,q]),'v':np.stack([v,v,v1]),'scale':np.ones(92),'a':np.stack([np.full(44,.1),np.full(44,.2),np.full(44,.3)]),'color_indices':ci,'dv':np.zeros((3,32))}
15
+ c.reset_episode();assert np.allclose(c.act(o),.1)
16
+ o['proprioception'][c._features.PATHS[0]][0]=3;c.reset_episode();assert np.allclose(c.act(o),.2)
17
+ o=obs();original=c._features.visual;c._features.visual=lambda rgb:v1;c.reset_episode();assert np.allclose(c.act(o),.3);c._features.visual=original
18
+ # Actual packaged memory with wholly synthetic input, without accuracy claims.
19
+ c._memory=None;c.reset_episode();o=obs();a=c.act(o);assert a.shape==(44,) and a.dtype==np.float32 and np.isfinite(a).all() and (abs(a)<=1).all()
20
+ o['step_id']=1499;assert c.act(o).shape==(44,)
21
+ o['step_id']=1500;assert c.act(o) is None
22
+ c.reset_episode();assert np.array_equal(c.act({'step_id':1}),np.zeros(44))
23
+ print('PASS: syntax/import, synthetic RGB formats, modality dependence, bounds, reset, missing input, horizon accounting. No environment tests.')
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/checks.txt ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ PASS: py_compile for controller.py, features.py, build_memory.py, check_synthetic.py.
2
+ PASS: synthetic HWC and 1HWC input, shape (44,), finite float32 actions, bounds [-1,1].
3
+ PASS: controlled synthetic memory selections change with proprioception and visual features.
4
+ PASS: reset, missing-input zero action, step 1499 action, step 1500 None.
5
+ PASS: packaged memory integrity: 9659 labels, 20 distinct authorized demo groups, finite arrays, legal action bounds.
6
+ These are syntax, synthetic-input and offline data integrity checks only. No task accuracy, rollout, rehearsal, success, or runtime frequency claim.
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/config.json ADDED
@@ -0,0 +1,130 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_parts": [
3
+ {
4
+ "robot_key": "panda_leaphand-1",
5
+ "width": 22
6
+ },
7
+ {
8
+ "robot_key": "panda_leaphand_left-0",
9
+ "width": 22
10
+ }
11
+ ],
12
+ "action_low": [
13
+ -1.0,
14
+ -1.0,
15
+ -1.0,
16
+ -1.0,
17
+ -1.0,
18
+ -1.0,
19
+ -1.0,
20
+ -1.0,
21
+ -1.0,
22
+ -1.0,
23
+ -1.0,
24
+ -1.0,
25
+ -1.0,
26
+ -1.0,
27
+ -1.0,
28
+ -1.0,
29
+ -1.0,
30
+ -1.0,
31
+ -1.0,
32
+ -1.0,
33
+ -1.0,
34
+ -1.0,
35
+ -1.0,
36
+ -1.0,
37
+ -1.0,
38
+ -1.0,
39
+ -1.0,
40
+ -1.0,
41
+ -1.0,
42
+ -1.0,
43
+ -1.0,
44
+ -1.0,
45
+ -1.0,
46
+ -1.0,
47
+ -1.0,
48
+ -1.0,
49
+ -1.0,
50
+ -1.0,
51
+ -1.0,
52
+ -1.0,
53
+ -1.0,
54
+ -1.0,
55
+ -1.0,
56
+ -1.0
57
+ ],
58
+ "action_high": [
59
+ 1.0,
60
+ 1.0,
61
+ 1.0,
62
+ 1.0,
63
+ 1.0,
64
+ 1.0,
65
+ 1.0,
66
+ 1.0,
67
+ 1.0,
68
+ 1.0,
69
+ 1.0,
70
+ 1.0,
71
+ 1.0,
72
+ 1.0,
73
+ 1.0,
74
+ 1.0,
75
+ 1.0,
76
+ 1.0,
77
+ 1.0,
78
+ 1.0,
79
+ 1.0,
80
+ 1.0,
81
+ 1.0,
82
+ 1.0,
83
+ 1.0,
84
+ 1.0,
85
+ 1.0,
86
+ 1.0,
87
+ 1.0,
88
+ 1.0,
89
+ 1.0,
90
+ 1.0,
91
+ 1.0,
92
+ 1.0,
93
+ 1.0,
94
+ 1.0,
95
+ 1.0,
96
+ 1.0,
97
+ 1.0,
98
+ 1.0,
99
+ 1.0,
100
+ 1.0,
101
+ 1.0,
102
+ 1.0
103
+ ],
104
+ "native_horizon": 1500,
105
+ "control_hz": 20,
106
+ "demo_ids": [
107
+ "demo_00",
108
+ "demo_01",
109
+ "demo_02",
110
+ "demo_03",
111
+ "demo_04",
112
+ "demo_05",
113
+ "demo_06",
114
+ "demo_07",
115
+ "demo_08",
116
+ "demo_09",
117
+ "demo_10",
118
+ "demo_11",
119
+ "demo_12",
120
+ "demo_13",
121
+ "demo_14",
122
+ "demo_15",
123
+ "demo_16",
124
+ "demo_17",
125
+ "demo_18",
126
+ "demo_19"
127
+ ],
128
+ "action_gain": 1,
129
+ "slow_after_frame": 300
130
+ }
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/controller.py ADDED
@@ -0,0 +1,68 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Untested D20 observation-conditioned nearest-example controller.
2
+ No simulator, hidden state, episode identity, learned neural weights, or time replay.
3
+ """
4
+ from pathlib import Path
5
+ import importlib.util
6
+ import json
7
+ import numpy as np
8
+ _ROOT=Path(__file__).resolve().parent
9
+ _spec=importlib.util.spec_from_file_location('_d20_features',_ROOT/'features.py')
10
+ _features=importlib.util.module_from_spec(_spec);_spec.loader.exec_module(_features)
11
+ _cfg=json.loads((_ROOT/'config.json').read_text())
12
+ _memory=None
13
+ _previous_visual=None
14
+ _last_step=None
15
+
16
+ def reset_episode():
17
+ global _previous_visual,_last_step
18
+ _previous_visual=None
19
+ _last_step=None
20
+
21
+ def _load():
22
+ global _memory
23
+ if _memory is None:
24
+ with np.load(_ROOT/'memory.npz',allow_pickle=False) as z:
25
+ _memory={k:z[k] for k in z.files}
26
+ v=_memory['v']
27
+ # Color summaries occupy first 16 entries of each 592-wide camera block.
28
+ ci=np.r_[np.arange(16),np.arange(592,608)]
29
+ _memory['color_indices']=ci
30
+ dv=np.zeros((len(v),32),np.float32)
31
+ same=_memory['demo'][1:]==_memory['demo'][:-1]
32
+ dv[1:]=np.where(same[:,None],v[1:,ci]-v[:-1,ci],0)
33
+ _memory['dv']=dv
34
+ return _memory
35
+
36
+ def act(obs):
37
+ global _previous_visual,_last_step,diagnostic
38
+ step=int(obs.get('step_id',1 if _last_step is None else _last_step+1))
39
+ if step>=_cfg['native_horizon']:
40
+ return None # 1 initial runner hold + at most 1499 policy actions.
41
+ if _last_step is not None and step<=_last_step:
42
+ reset_episode()
43
+ try:
44
+ q=_features.proprio(obs['proprioception'])
45
+ v=_features.visual(obs['rgb'])
46
+ if not np.isfinite(v).all(): raise ValueError('Nonfinite RGB features')
47
+ except (KeyError,ValueError,TypeError):
48
+ _previous_visual=None
49
+ _last_step=step
50
+ return np.zeros(44,np.float32)
51
+ m=_load();ci=m['color_indices']
52
+ qdist=np.mean(((m['q']-q)/m['scale'])**2,axis=1)
53
+ # Resized RGB preserves hand geometry; colored components emphasize balls.
54
+ vdist=np.mean((m['v']-v)**2,axis=1)
55
+ cdist=np.mean((m['v'][:,ci]-v[ci])**2,axis=1)
56
+ score=qdist+_cfg.get('visual_weight',4)*vdist+_cfg.get('color_weight',2)*cdist
57
+ if _previous_visual is not None and _last_step==step-1:
58
+ dv=v[ci]-_previous_visual[ci]
59
+ score+=.25*np.mean((m['dv']-dv)**2,axis=1)
60
+ j=int(np.argmin(score))
61
+ diagnostic={'demo':int(m['demo'][j]),'frame':int(m['frame'][j]),'qdist':float(qdist[j]),'vdist':float(vdist[j])}
62
+ _previous_visual=v.copy();_last_step=step
63
+ if qdist[j]>25 or vdist[j]>.25:
64
+ return np.zeros(44,np.float32) # Conservative, uncalibrated OOD hold.
65
+ # Label was concatenated in contract.action_parts order during compilation.
66
+ gain=_cfg.get('action_gain',1)
67
+ if int(m['frame'][j])>=_cfg.get('slow_after_frame',10000): gain=1.0
68
+ return np.clip(m['a'][j]*gain,_cfg['action_low'],_cfg['action_high']).astype(np.float32)
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/demo_manifest.json ADDED
@@ -0,0 +1,2168 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task_id": "DualArmJugglingBall-v1",
3
+ "status": "released",
4
+ "selection_rule": "prefer existing pinned D20, otherwise numeric trajectory ID sort before reading data",
5
+ "lineage_status": "pinned_comparator",
6
+ "episodes": [
7
+ {
8
+ "demo_id": "demo_00",
9
+ "file": "demo_00.h5",
10
+ "length": 487,
11
+ "fields": [
12
+ {
13
+ "path": "actions/panda_leaphand-1",
14
+ "shape": [
15
+ 487,
16
+ 22
17
+ ],
18
+ "dtype": "float32"
19
+ },
20
+ {
21
+ "path": "actions/panda_leaphand_left-0",
22
+ "shape": [
23
+ 487,
24
+ 22
25
+ ],
26
+ "dtype": "float32"
27
+ },
28
+ {
29
+ "path": "obs/agent/panda_leaphand-1/qpos",
30
+ "shape": [
31
+ 488,
32
+ 23
33
+ ],
34
+ "dtype": "float32"
35
+ },
36
+ {
37
+ "path": "obs/agent/panda_leaphand-1/qvel",
38
+ "shape": [
39
+ 488,
40
+ 23
41
+ ],
42
+ "dtype": "float32"
43
+ },
44
+ {
45
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
46
+ "shape": [
47
+ 488,
48
+ 23
49
+ ],
50
+ "dtype": "float32"
51
+ },
52
+ {
53
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
54
+ "shape": [
55
+ 488,
56
+ 23
57
+ ],
58
+ "dtype": "float32"
59
+ },
60
+ {
61
+ "path": "obs/sensor_data/front_camera/rgb",
62
+ "shape": [
63
+ 488,
64
+ 240,
65
+ 320,
66
+ 3
67
+ ],
68
+ "dtype": "uint8"
69
+ },
70
+ {
71
+ "path": "obs/sensor_data/left_camera/rgb",
72
+ "shape": [
73
+ 488,
74
+ 240,
75
+ 320,
76
+ 3
77
+ ],
78
+ "dtype": "uint8"
79
+ },
80
+ {
81
+ "path": "obs/sensor_data/robot_camera_0/rgb",
82
+ "shape": [
83
+ 488,
84
+ 240,
85
+ 320,
86
+ 3
87
+ ],
88
+ "dtype": "uint8"
89
+ },
90
+ {
91
+ "path": "obs/sensor_data/robot_camera_1/rgb",
92
+ "shape": [
93
+ 488,
94
+ 240,
95
+ 320,
96
+ 3
97
+ ],
98
+ "dtype": "uint8"
99
+ },
100
+ {
101
+ "path": "obs/sensor_data/top_camera/rgb",
102
+ "shape": [
103
+ 488,
104
+ 320,
105
+ 240,
106
+ 3
107
+ ],
108
+ "dtype": "uint8"
109
+ }
110
+ ],
111
+ "trajectory_sha256": "3c8802a3139066326eb76b0e02ad8a8e078e590892ac64749d01a672c2dd50d1",
112
+ "source_sha256": "55a9c8935926c83051b207d9a764214c0dda40db7c05a5e894363127207feb7c",
113
+ "view_sha256": "e6788fbf5e51d0d75bb3c0b9b63f6e0deea78cc72826ef6b7adb1dfdc9a866cd"
114
+ },
115
+ {
116
+ "demo_id": "demo_01",
117
+ "file": "demo_01.h5",
118
+ "length": 481,
119
+ "fields": [
120
+ {
121
+ "path": "actions/panda_leaphand-1",
122
+ "shape": [
123
+ 481,
124
+ 22
125
+ ],
126
+ "dtype": "float32"
127
+ },
128
+ {
129
+ "path": "actions/panda_leaphand_left-0",
130
+ "shape": [
131
+ 481,
132
+ 22
133
+ ],
134
+ "dtype": "float32"
135
+ },
136
+ {
137
+ "path": "obs/agent/panda_leaphand-1/qpos",
138
+ "shape": [
139
+ 482,
140
+ 23
141
+ ],
142
+ "dtype": "float32"
143
+ },
144
+ {
145
+ "path": "obs/agent/panda_leaphand-1/qvel",
146
+ "shape": [
147
+ 482,
148
+ 23
149
+ ],
150
+ "dtype": "float32"
151
+ },
152
+ {
153
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
154
+ "shape": [
155
+ 482,
156
+ 23
157
+ ],
158
+ "dtype": "float32"
159
+ },
160
+ {
161
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
162
+ "shape": [
163
+ 482,
164
+ 23
165
+ ],
166
+ "dtype": "float32"
167
+ },
168
+ {
169
+ "path": "obs/sensor_data/front_camera/rgb",
170
+ "shape": [
171
+ 482,
172
+ 240,
173
+ 320,
174
+ 3
175
+ ],
176
+ "dtype": "uint8"
177
+ },
178
+ {
179
+ "path": "obs/sensor_data/left_camera/rgb",
180
+ "shape": [
181
+ 482,
182
+ 240,
183
+ 320,
184
+ 3
185
+ ],
186
+ "dtype": "uint8"
187
+ },
188
+ {
189
+ "path": "obs/sensor_data/robot_camera_0/rgb",
190
+ "shape": [
191
+ 482,
192
+ 240,
193
+ 320,
194
+ 3
195
+ ],
196
+ "dtype": "uint8"
197
+ },
198
+ {
199
+ "path": "obs/sensor_data/robot_camera_1/rgb",
200
+ "shape": [
201
+ 482,
202
+ 240,
203
+ 320,
204
+ 3
205
+ ],
206
+ "dtype": "uint8"
207
+ },
208
+ {
209
+ "path": "obs/sensor_data/top_camera/rgb",
210
+ "shape": [
211
+ 482,
212
+ 320,
213
+ 240,
214
+ 3
215
+ ],
216
+ "dtype": "uint8"
217
+ }
218
+ ],
219
+ "trajectory_sha256": "a2de3d380116c08ad81a02291375b09a42c54f523fd405a52efdcf19cfd2ef13",
220
+ "source_sha256": "7bce670de50b1d6b0a88fedf5faac0bee3ab7644482d31240f9d953e04ca7722",
221
+ "view_sha256": "e5c5ccc5750e50aa84d3a162c2d440bc107c700caaf94dc5baf1b8a10da3baff"
222
+ },
223
+ {
224
+ "demo_id": "demo_02",
225
+ "file": "demo_02.h5",
226
+ "length": 486,
227
+ "fields": [
228
+ {
229
+ "path": "actions/panda_leaphand-1",
230
+ "shape": [
231
+ 486,
232
+ 22
233
+ ],
234
+ "dtype": "float32"
235
+ },
236
+ {
237
+ "path": "actions/panda_leaphand_left-0",
238
+ "shape": [
239
+ 486,
240
+ 22
241
+ ],
242
+ "dtype": "float32"
243
+ },
244
+ {
245
+ "path": "obs/agent/panda_leaphand-1/qpos",
246
+ "shape": [
247
+ 487,
248
+ 23
249
+ ],
250
+ "dtype": "float32"
251
+ },
252
+ {
253
+ "path": "obs/agent/panda_leaphand-1/qvel",
254
+ "shape": [
255
+ 487,
256
+ 23
257
+ ],
258
+ "dtype": "float32"
259
+ },
260
+ {
261
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
262
+ "shape": [
263
+ 487,
264
+ 23
265
+ ],
266
+ "dtype": "float32"
267
+ },
268
+ {
269
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
270
+ "shape": [
271
+ 487,
272
+ 23
273
+ ],
274
+ "dtype": "float32"
275
+ },
276
+ {
277
+ "path": "obs/sensor_data/front_camera/rgb",
278
+ "shape": [
279
+ 487,
280
+ 240,
281
+ 320,
282
+ 3
283
+ ],
284
+ "dtype": "uint8"
285
+ },
286
+ {
287
+ "path": "obs/sensor_data/left_camera/rgb",
288
+ "shape": [
289
+ 487,
290
+ 240,
291
+ 320,
292
+ 3
293
+ ],
294
+ "dtype": "uint8"
295
+ },
296
+ {
297
+ "path": "obs/sensor_data/robot_camera_0/rgb",
298
+ "shape": [
299
+ 487,
300
+ 240,
301
+ 320,
302
+ 3
303
+ ],
304
+ "dtype": "uint8"
305
+ },
306
+ {
307
+ "path": "obs/sensor_data/robot_camera_1/rgb",
308
+ "shape": [
309
+ 487,
310
+ 240,
311
+ 320,
312
+ 3
313
+ ],
314
+ "dtype": "uint8"
315
+ },
316
+ {
317
+ "path": "obs/sensor_data/top_camera/rgb",
318
+ "shape": [
319
+ 487,
320
+ 320,
321
+ 240,
322
+ 3
323
+ ],
324
+ "dtype": "uint8"
325
+ }
326
+ ],
327
+ "trajectory_sha256": "7ffb1f1c783af07c19ee8cc0778483d51a3ab57ce77fe781ff6fa52b6d5a825e",
328
+ "source_sha256": "5e080dcd5b26f1ae447938d74ec521ae47d6ae2b84c3902c64bb38e7f82a3eeb",
329
+ "view_sha256": "4eeb1b7f8ea2f08042656412359036eb1c62f7388240507bc5455947ac156a23"
330
+ },
331
+ {
332
+ "demo_id": "demo_03",
333
+ "file": "demo_03.h5",
334
+ "length": 478,
335
+ "fields": [
336
+ {
337
+ "path": "actions/panda_leaphand-1",
338
+ "shape": [
339
+ 478,
340
+ 22
341
+ ],
342
+ "dtype": "float32"
343
+ },
344
+ {
345
+ "path": "actions/panda_leaphand_left-0",
346
+ "shape": [
347
+ 478,
348
+ 22
349
+ ],
350
+ "dtype": "float32"
351
+ },
352
+ {
353
+ "path": "obs/agent/panda_leaphand-1/qpos",
354
+ "shape": [
355
+ 479,
356
+ 23
357
+ ],
358
+ "dtype": "float32"
359
+ },
360
+ {
361
+ "path": "obs/agent/panda_leaphand-1/qvel",
362
+ "shape": [
363
+ 479,
364
+ 23
365
+ ],
366
+ "dtype": "float32"
367
+ },
368
+ {
369
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
370
+ "shape": [
371
+ 479,
372
+ 23
373
+ ],
374
+ "dtype": "float32"
375
+ },
376
+ {
377
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
378
+ "shape": [
379
+ 479,
380
+ 23
381
+ ],
382
+ "dtype": "float32"
383
+ },
384
+ {
385
+ "path": "obs/sensor_data/front_camera/rgb",
386
+ "shape": [
387
+ 479,
388
+ 240,
389
+ 320,
390
+ 3
391
+ ],
392
+ "dtype": "uint8"
393
+ },
394
+ {
395
+ "path": "obs/sensor_data/left_camera/rgb",
396
+ "shape": [
397
+ 479,
398
+ 240,
399
+ 320,
400
+ 3
401
+ ],
402
+ "dtype": "uint8"
403
+ },
404
+ {
405
+ "path": "obs/sensor_data/robot_camera_0/rgb",
406
+ "shape": [
407
+ 479,
408
+ 240,
409
+ 320,
410
+ 3
411
+ ],
412
+ "dtype": "uint8"
413
+ },
414
+ {
415
+ "path": "obs/sensor_data/robot_camera_1/rgb",
416
+ "shape": [
417
+ 479,
418
+ 240,
419
+ 320,
420
+ 3
421
+ ],
422
+ "dtype": "uint8"
423
+ },
424
+ {
425
+ "path": "obs/sensor_data/top_camera/rgb",
426
+ "shape": [
427
+ 479,
428
+ 320,
429
+ 240,
430
+ 3
431
+ ],
432
+ "dtype": "uint8"
433
+ }
434
+ ],
435
+ "trajectory_sha256": "ebb767a8d84444fc7b4b776155f90144fd5505dfce6935dd35b7b82e242f0111",
436
+ "source_sha256": "b37e793b5ab1c106ab62e2095eb4f508efa3f3680af69d34b1238db77bfadaf8",
437
+ "view_sha256": "d56aa11e10fe1dede2e9b29118ff375e9622cb427fc7d41e314b8357979a03ad"
438
+ },
439
+ {
440
+ "demo_id": "demo_04",
441
+ "file": "demo_04.h5",
442
+ "length": 485,
443
+ "fields": [
444
+ {
445
+ "path": "actions/panda_leaphand-1",
446
+ "shape": [
447
+ 485,
448
+ 22
449
+ ],
450
+ "dtype": "float32"
451
+ },
452
+ {
453
+ "path": "actions/panda_leaphand_left-0",
454
+ "shape": [
455
+ 485,
456
+ 22
457
+ ],
458
+ "dtype": "float32"
459
+ },
460
+ {
461
+ "path": "obs/agent/panda_leaphand-1/qpos",
462
+ "shape": [
463
+ 486,
464
+ 23
465
+ ],
466
+ "dtype": "float32"
467
+ },
468
+ {
469
+ "path": "obs/agent/panda_leaphand-1/qvel",
470
+ "shape": [
471
+ 486,
472
+ 23
473
+ ],
474
+ "dtype": "float32"
475
+ },
476
+ {
477
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
478
+ "shape": [
479
+ 486,
480
+ 23
481
+ ],
482
+ "dtype": "float32"
483
+ },
484
+ {
485
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
486
+ "shape": [
487
+ 486,
488
+ 23
489
+ ],
490
+ "dtype": "float32"
491
+ },
492
+ {
493
+ "path": "obs/sensor_data/front_camera/rgb",
494
+ "shape": [
495
+ 486,
496
+ 240,
497
+ 320,
498
+ 3
499
+ ],
500
+ "dtype": "uint8"
501
+ },
502
+ {
503
+ "path": "obs/sensor_data/left_camera/rgb",
504
+ "shape": [
505
+ 486,
506
+ 240,
507
+ 320,
508
+ 3
509
+ ],
510
+ "dtype": "uint8"
511
+ },
512
+ {
513
+ "path": "obs/sensor_data/robot_camera_0/rgb",
514
+ "shape": [
515
+ 486,
516
+ 240,
517
+ 320,
518
+ 3
519
+ ],
520
+ "dtype": "uint8"
521
+ },
522
+ {
523
+ "path": "obs/sensor_data/robot_camera_1/rgb",
524
+ "shape": [
525
+ 486,
526
+ 240,
527
+ 320,
528
+ 3
529
+ ],
530
+ "dtype": "uint8"
531
+ },
532
+ {
533
+ "path": "obs/sensor_data/top_camera/rgb",
534
+ "shape": [
535
+ 486,
536
+ 320,
537
+ 240,
538
+ 3
539
+ ],
540
+ "dtype": "uint8"
541
+ }
542
+ ],
543
+ "trajectory_sha256": "23e581809a8f41e8c900e8f9f87c36caee76947ccc8c03f1b7ca970bd5a528f7",
544
+ "source_sha256": "21e54f62497f13d59648eb23cee8ff131e79fddc5c0c245fe49933fe7408f074",
545
+ "view_sha256": "2f3b321cdc749997bf4906e4022684bdd0fe2a4e3773c4bbea740313a410b652"
546
+ },
547
+ {
548
+ "demo_id": "demo_05",
549
+ "file": "demo_05.h5",
550
+ "length": 483,
551
+ "fields": [
552
+ {
553
+ "path": "actions/panda_leaphand-1",
554
+ "shape": [
555
+ 483,
556
+ 22
557
+ ],
558
+ "dtype": "float32"
559
+ },
560
+ {
561
+ "path": "actions/panda_leaphand_left-0",
562
+ "shape": [
563
+ 483,
564
+ 22
565
+ ],
566
+ "dtype": "float32"
567
+ },
568
+ {
569
+ "path": "obs/agent/panda_leaphand-1/qpos",
570
+ "shape": [
571
+ 484,
572
+ 23
573
+ ],
574
+ "dtype": "float32"
575
+ },
576
+ {
577
+ "path": "obs/agent/panda_leaphand-1/qvel",
578
+ "shape": [
579
+ 484,
580
+ 23
581
+ ],
582
+ "dtype": "float32"
583
+ },
584
+ {
585
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
586
+ "shape": [
587
+ 484,
588
+ 23
589
+ ],
590
+ "dtype": "float32"
591
+ },
592
+ {
593
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
594
+ "shape": [
595
+ 484,
596
+ 23
597
+ ],
598
+ "dtype": "float32"
599
+ },
600
+ {
601
+ "path": "obs/sensor_data/front_camera/rgb",
602
+ "shape": [
603
+ 484,
604
+ 240,
605
+ 320,
606
+ 3
607
+ ],
608
+ "dtype": "uint8"
609
+ },
610
+ {
611
+ "path": "obs/sensor_data/left_camera/rgb",
612
+ "shape": [
613
+ 484,
614
+ 240,
615
+ 320,
616
+ 3
617
+ ],
618
+ "dtype": "uint8"
619
+ },
620
+ {
621
+ "path": "obs/sensor_data/robot_camera_0/rgb",
622
+ "shape": [
623
+ 484,
624
+ 240,
625
+ 320,
626
+ 3
627
+ ],
628
+ "dtype": "uint8"
629
+ },
630
+ {
631
+ "path": "obs/sensor_data/robot_camera_1/rgb",
632
+ "shape": [
633
+ 484,
634
+ 240,
635
+ 320,
636
+ 3
637
+ ],
638
+ "dtype": "uint8"
639
+ },
640
+ {
641
+ "path": "obs/sensor_data/top_camera/rgb",
642
+ "shape": [
643
+ 484,
644
+ 320,
645
+ 240,
646
+ 3
647
+ ],
648
+ "dtype": "uint8"
649
+ }
650
+ ],
651
+ "trajectory_sha256": "a83e1f09da6f072a814825bc97a0579da777b0389f3d7d1448079b0633caa363",
652
+ "source_sha256": "75bd092f088b856391c38cf05ca15170eaa2252231b42a17012713b728910974",
653
+ "view_sha256": "a654d095a6773246ce04c021941f36800ac984e8d394252c3ff392e446bb2242"
654
+ },
655
+ {
656
+ "demo_id": "demo_06",
657
+ "file": "demo_06.h5",
658
+ "length": 478,
659
+ "fields": [
660
+ {
661
+ "path": "actions/panda_leaphand-1",
662
+ "shape": [
663
+ 478,
664
+ 22
665
+ ],
666
+ "dtype": "float32"
667
+ },
668
+ {
669
+ "path": "actions/panda_leaphand_left-0",
670
+ "shape": [
671
+ 478,
672
+ 22
673
+ ],
674
+ "dtype": "float32"
675
+ },
676
+ {
677
+ "path": "obs/agent/panda_leaphand-1/qpos",
678
+ "shape": [
679
+ 479,
680
+ 23
681
+ ],
682
+ "dtype": "float32"
683
+ },
684
+ {
685
+ "path": "obs/agent/panda_leaphand-1/qvel",
686
+ "shape": [
687
+ 479,
688
+ 23
689
+ ],
690
+ "dtype": "float32"
691
+ },
692
+ {
693
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
694
+ "shape": [
695
+ 479,
696
+ 23
697
+ ],
698
+ "dtype": "float32"
699
+ },
700
+ {
701
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
702
+ "shape": [
703
+ 479,
704
+ 23
705
+ ],
706
+ "dtype": "float32"
707
+ },
708
+ {
709
+ "path": "obs/sensor_data/front_camera/rgb",
710
+ "shape": [
711
+ 479,
712
+ 240,
713
+ 320,
714
+ 3
715
+ ],
716
+ "dtype": "uint8"
717
+ },
718
+ {
719
+ "path": "obs/sensor_data/left_camera/rgb",
720
+ "shape": [
721
+ 479,
722
+ 240,
723
+ 320,
724
+ 3
725
+ ],
726
+ "dtype": "uint8"
727
+ },
728
+ {
729
+ "path": "obs/sensor_data/robot_camera_0/rgb",
730
+ "shape": [
731
+ 479,
732
+ 240,
733
+ 320,
734
+ 3
735
+ ],
736
+ "dtype": "uint8"
737
+ },
738
+ {
739
+ "path": "obs/sensor_data/robot_camera_1/rgb",
740
+ "shape": [
741
+ 479,
742
+ 240,
743
+ 320,
744
+ 3
745
+ ],
746
+ "dtype": "uint8"
747
+ },
748
+ {
749
+ "path": "obs/sensor_data/top_camera/rgb",
750
+ "shape": [
751
+ 479,
752
+ 320,
753
+ 240,
754
+ 3
755
+ ],
756
+ "dtype": "uint8"
757
+ }
758
+ ],
759
+ "trajectory_sha256": "994e307d189f770049d912573d00acc101c83dc8fc39064cb62669e431167d0a",
760
+ "source_sha256": "0a2f4d3728d6506965f199ae72c4d4d37bb676a4513b56fbe759c7ea5991e50a",
761
+ "view_sha256": "a318cc09e376a66ec514df7b215bc12b5ae15ece64fe4bbe39df3a33252e1dfd"
762
+ },
763
+ {
764
+ "demo_id": "demo_07",
765
+ "file": "demo_07.h5",
766
+ "length": 483,
767
+ "fields": [
768
+ {
769
+ "path": "actions/panda_leaphand-1",
770
+ "shape": [
771
+ 483,
772
+ 22
773
+ ],
774
+ "dtype": "float32"
775
+ },
776
+ {
777
+ "path": "actions/panda_leaphand_left-0",
778
+ "shape": [
779
+ 483,
780
+ 22
781
+ ],
782
+ "dtype": "float32"
783
+ },
784
+ {
785
+ "path": "obs/agent/panda_leaphand-1/qpos",
786
+ "shape": [
787
+ 484,
788
+ 23
789
+ ],
790
+ "dtype": "float32"
791
+ },
792
+ {
793
+ "path": "obs/agent/panda_leaphand-1/qvel",
794
+ "shape": [
795
+ 484,
796
+ 23
797
+ ],
798
+ "dtype": "float32"
799
+ },
800
+ {
801
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
802
+ "shape": [
803
+ 484,
804
+ 23
805
+ ],
806
+ "dtype": "float32"
807
+ },
808
+ {
809
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
810
+ "shape": [
811
+ 484,
812
+ 23
813
+ ],
814
+ "dtype": "float32"
815
+ },
816
+ {
817
+ "path": "obs/sensor_data/front_camera/rgb",
818
+ "shape": [
819
+ 484,
820
+ 240,
821
+ 320,
822
+ 3
823
+ ],
824
+ "dtype": "uint8"
825
+ },
826
+ {
827
+ "path": "obs/sensor_data/left_camera/rgb",
828
+ "shape": [
829
+ 484,
830
+ 240,
831
+ 320,
832
+ 3
833
+ ],
834
+ "dtype": "uint8"
835
+ },
836
+ {
837
+ "path": "obs/sensor_data/robot_camera_0/rgb",
838
+ "shape": [
839
+ 484,
840
+ 240,
841
+ 320,
842
+ 3
843
+ ],
844
+ "dtype": "uint8"
845
+ },
846
+ {
847
+ "path": "obs/sensor_data/robot_camera_1/rgb",
848
+ "shape": [
849
+ 484,
850
+ 240,
851
+ 320,
852
+ 3
853
+ ],
854
+ "dtype": "uint8"
855
+ },
856
+ {
857
+ "path": "obs/sensor_data/top_camera/rgb",
858
+ "shape": [
859
+ 484,
860
+ 320,
861
+ 240,
862
+ 3
863
+ ],
864
+ "dtype": "uint8"
865
+ }
866
+ ],
867
+ "trajectory_sha256": "d0618f2f8b45c4b2c378b61b446676f65c08838b7460640be18e51a53f4860ff",
868
+ "source_sha256": "fc88989e6b908dca7803dc988298f19110a3331e6ed04c1c1169b111b83c504e",
869
+ "view_sha256": "1e78e3399a4e6d0a6f71547ade517cc428d39c28b5e01a2690550fa3128f51ef"
870
+ },
871
+ {
872
+ "demo_id": "demo_08",
873
+ "file": "demo_08.h5",
874
+ "length": 483,
875
+ "fields": [
876
+ {
877
+ "path": "actions/panda_leaphand-1",
878
+ "shape": [
879
+ 483,
880
+ 22
881
+ ],
882
+ "dtype": "float32"
883
+ },
884
+ {
885
+ "path": "actions/panda_leaphand_left-0",
886
+ "shape": [
887
+ 483,
888
+ 22
889
+ ],
890
+ "dtype": "float32"
891
+ },
892
+ {
893
+ "path": "obs/agent/panda_leaphand-1/qpos",
894
+ "shape": [
895
+ 484,
896
+ 23
897
+ ],
898
+ "dtype": "float32"
899
+ },
900
+ {
901
+ "path": "obs/agent/panda_leaphand-1/qvel",
902
+ "shape": [
903
+ 484,
904
+ 23
905
+ ],
906
+ "dtype": "float32"
907
+ },
908
+ {
909
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
910
+ "shape": [
911
+ 484,
912
+ 23
913
+ ],
914
+ "dtype": "float32"
915
+ },
916
+ {
917
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
918
+ "shape": [
919
+ 484,
920
+ 23
921
+ ],
922
+ "dtype": "float32"
923
+ },
924
+ {
925
+ "path": "obs/sensor_data/front_camera/rgb",
926
+ "shape": [
927
+ 484,
928
+ 240,
929
+ 320,
930
+ 3
931
+ ],
932
+ "dtype": "uint8"
933
+ },
934
+ {
935
+ "path": "obs/sensor_data/left_camera/rgb",
936
+ "shape": [
937
+ 484,
938
+ 240,
939
+ 320,
940
+ 3
941
+ ],
942
+ "dtype": "uint8"
943
+ },
944
+ {
945
+ "path": "obs/sensor_data/robot_camera_0/rgb",
946
+ "shape": [
947
+ 484,
948
+ 240,
949
+ 320,
950
+ 3
951
+ ],
952
+ "dtype": "uint8"
953
+ },
954
+ {
955
+ "path": "obs/sensor_data/robot_camera_1/rgb",
956
+ "shape": [
957
+ 484,
958
+ 240,
959
+ 320,
960
+ 3
961
+ ],
962
+ "dtype": "uint8"
963
+ },
964
+ {
965
+ "path": "obs/sensor_data/top_camera/rgb",
966
+ "shape": [
967
+ 484,
968
+ 320,
969
+ 240,
970
+ 3
971
+ ],
972
+ "dtype": "uint8"
973
+ }
974
+ ],
975
+ "trajectory_sha256": "1dca0383cc799d05bb5e3af8c9e29cc8ddae24c46d0c5c12a52a0a2b7ee28871",
976
+ "source_sha256": "7a1f9df1e78d215fdc6b299bd60807a606e4d3ba9909c1dbbe95df3d282ded78",
977
+ "view_sha256": "61e858727b4ea4e2eeff1c4e156c575f39beaeeede80100f5906c98700623d6b"
978
+ },
979
+ {
980
+ "demo_id": "demo_09",
981
+ "file": "demo_09.h5",
982
+ "length": 485,
983
+ "fields": [
984
+ {
985
+ "path": "actions/panda_leaphand-1",
986
+ "shape": [
987
+ 485,
988
+ 22
989
+ ],
990
+ "dtype": "float32"
991
+ },
992
+ {
993
+ "path": "actions/panda_leaphand_left-0",
994
+ "shape": [
995
+ 485,
996
+ 22
997
+ ],
998
+ "dtype": "float32"
999
+ },
1000
+ {
1001
+ "path": "obs/agent/panda_leaphand-1/qpos",
1002
+ "shape": [
1003
+ 486,
1004
+ 23
1005
+ ],
1006
+ "dtype": "float32"
1007
+ },
1008
+ {
1009
+ "path": "obs/agent/panda_leaphand-1/qvel",
1010
+ "shape": [
1011
+ 486,
1012
+ 23
1013
+ ],
1014
+ "dtype": "float32"
1015
+ },
1016
+ {
1017
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
1018
+ "shape": [
1019
+ 486,
1020
+ 23
1021
+ ],
1022
+ "dtype": "float32"
1023
+ },
1024
+ {
1025
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
1026
+ "shape": [
1027
+ 486,
1028
+ 23
1029
+ ],
1030
+ "dtype": "float32"
1031
+ },
1032
+ {
1033
+ "path": "obs/sensor_data/front_camera/rgb",
1034
+ "shape": [
1035
+ 486,
1036
+ 240,
1037
+ 320,
1038
+ 3
1039
+ ],
1040
+ "dtype": "uint8"
1041
+ },
1042
+ {
1043
+ "path": "obs/sensor_data/left_camera/rgb",
1044
+ "shape": [
1045
+ 486,
1046
+ 240,
1047
+ 320,
1048
+ 3
1049
+ ],
1050
+ "dtype": "uint8"
1051
+ },
1052
+ {
1053
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1054
+ "shape": [
1055
+ 486,
1056
+ 240,
1057
+ 320,
1058
+ 3
1059
+ ],
1060
+ "dtype": "uint8"
1061
+ },
1062
+ {
1063
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1064
+ "shape": [
1065
+ 486,
1066
+ 240,
1067
+ 320,
1068
+ 3
1069
+ ],
1070
+ "dtype": "uint8"
1071
+ },
1072
+ {
1073
+ "path": "obs/sensor_data/top_camera/rgb",
1074
+ "shape": [
1075
+ 486,
1076
+ 320,
1077
+ 240,
1078
+ 3
1079
+ ],
1080
+ "dtype": "uint8"
1081
+ }
1082
+ ],
1083
+ "trajectory_sha256": "da0769c70bb9e521a6c99bd6109d0620a2b4d468def58a2359e160df05973fff",
1084
+ "source_sha256": "41a191be8cb0bf9901e51042f7bf79bb762102b7eaed5f57a03f7dd9ad3b6f97",
1085
+ "view_sha256": "11feb420b28449d0fb6b4119ca5ecccefe63bc7a6f0a52d7f0bbd7284d4bea4e"
1086
+ },
1087
+ {
1088
+ "demo_id": "demo_10",
1089
+ "file": "demo_10.h5",
1090
+ "length": 481,
1091
+ "fields": [
1092
+ {
1093
+ "path": "actions/panda_leaphand-1",
1094
+ "shape": [
1095
+ 481,
1096
+ 22
1097
+ ],
1098
+ "dtype": "float32"
1099
+ },
1100
+ {
1101
+ "path": "actions/panda_leaphand_left-0",
1102
+ "shape": [
1103
+ 481,
1104
+ 22
1105
+ ],
1106
+ "dtype": "float32"
1107
+ },
1108
+ {
1109
+ "path": "obs/agent/panda_leaphand-1/qpos",
1110
+ "shape": [
1111
+ 482,
1112
+ 23
1113
+ ],
1114
+ "dtype": "float32"
1115
+ },
1116
+ {
1117
+ "path": "obs/agent/panda_leaphand-1/qvel",
1118
+ "shape": [
1119
+ 482,
1120
+ 23
1121
+ ],
1122
+ "dtype": "float32"
1123
+ },
1124
+ {
1125
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
1126
+ "shape": [
1127
+ 482,
1128
+ 23
1129
+ ],
1130
+ "dtype": "float32"
1131
+ },
1132
+ {
1133
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
1134
+ "shape": [
1135
+ 482,
1136
+ 23
1137
+ ],
1138
+ "dtype": "float32"
1139
+ },
1140
+ {
1141
+ "path": "obs/sensor_data/front_camera/rgb",
1142
+ "shape": [
1143
+ 482,
1144
+ 240,
1145
+ 320,
1146
+ 3
1147
+ ],
1148
+ "dtype": "uint8"
1149
+ },
1150
+ {
1151
+ "path": "obs/sensor_data/left_camera/rgb",
1152
+ "shape": [
1153
+ 482,
1154
+ 240,
1155
+ 320,
1156
+ 3
1157
+ ],
1158
+ "dtype": "uint8"
1159
+ },
1160
+ {
1161
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1162
+ "shape": [
1163
+ 482,
1164
+ 240,
1165
+ 320,
1166
+ 3
1167
+ ],
1168
+ "dtype": "uint8"
1169
+ },
1170
+ {
1171
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1172
+ "shape": [
1173
+ 482,
1174
+ 240,
1175
+ 320,
1176
+ 3
1177
+ ],
1178
+ "dtype": "uint8"
1179
+ },
1180
+ {
1181
+ "path": "obs/sensor_data/top_camera/rgb",
1182
+ "shape": [
1183
+ 482,
1184
+ 320,
1185
+ 240,
1186
+ 3
1187
+ ],
1188
+ "dtype": "uint8"
1189
+ }
1190
+ ],
1191
+ "trajectory_sha256": "eb1f28540074dc85d84cbf8ce3c201ebff6e08371e56c84071554e15f7a5c459",
1192
+ "source_sha256": "6d4b765cf67d236f3e604d5f536704ed107e52bf8d15207efed00d34f09e8587",
1193
+ "view_sha256": "f9914be698a1caebd2b56f34a34a6a808b1f63c9c1328ec54290e2f43fe2d219"
1194
+ },
1195
+ {
1196
+ "demo_id": "demo_11",
1197
+ "file": "demo_11.h5",
1198
+ "length": 483,
1199
+ "fields": [
1200
+ {
1201
+ "path": "actions/panda_leaphand-1",
1202
+ "shape": [
1203
+ 483,
1204
+ 22
1205
+ ],
1206
+ "dtype": "float32"
1207
+ },
1208
+ {
1209
+ "path": "actions/panda_leaphand_left-0",
1210
+ "shape": [
1211
+ 483,
1212
+ 22
1213
+ ],
1214
+ "dtype": "float32"
1215
+ },
1216
+ {
1217
+ "path": "obs/agent/panda_leaphand-1/qpos",
1218
+ "shape": [
1219
+ 484,
1220
+ 23
1221
+ ],
1222
+ "dtype": "float32"
1223
+ },
1224
+ {
1225
+ "path": "obs/agent/panda_leaphand-1/qvel",
1226
+ "shape": [
1227
+ 484,
1228
+ 23
1229
+ ],
1230
+ "dtype": "float32"
1231
+ },
1232
+ {
1233
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
1234
+ "shape": [
1235
+ 484,
1236
+ 23
1237
+ ],
1238
+ "dtype": "float32"
1239
+ },
1240
+ {
1241
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
1242
+ "shape": [
1243
+ 484,
1244
+ 23
1245
+ ],
1246
+ "dtype": "float32"
1247
+ },
1248
+ {
1249
+ "path": "obs/sensor_data/front_camera/rgb",
1250
+ "shape": [
1251
+ 484,
1252
+ 240,
1253
+ 320,
1254
+ 3
1255
+ ],
1256
+ "dtype": "uint8"
1257
+ },
1258
+ {
1259
+ "path": "obs/sensor_data/left_camera/rgb",
1260
+ "shape": [
1261
+ 484,
1262
+ 240,
1263
+ 320,
1264
+ 3
1265
+ ],
1266
+ "dtype": "uint8"
1267
+ },
1268
+ {
1269
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1270
+ "shape": [
1271
+ 484,
1272
+ 240,
1273
+ 320,
1274
+ 3
1275
+ ],
1276
+ "dtype": "uint8"
1277
+ },
1278
+ {
1279
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1280
+ "shape": [
1281
+ 484,
1282
+ 240,
1283
+ 320,
1284
+ 3
1285
+ ],
1286
+ "dtype": "uint8"
1287
+ },
1288
+ {
1289
+ "path": "obs/sensor_data/top_camera/rgb",
1290
+ "shape": [
1291
+ 484,
1292
+ 320,
1293
+ 240,
1294
+ 3
1295
+ ],
1296
+ "dtype": "uint8"
1297
+ }
1298
+ ],
1299
+ "trajectory_sha256": "97455d401afa30cb5653f867f5b0144e6adaa66864dd43b39c1c83b049990c7f",
1300
+ "source_sha256": "3a3ab1654322c575678dc63e9d9228e1638955745a0afb833330aa448f94a442",
1301
+ "view_sha256": "384997aba3ebf7d9942eed1af72cd67a7e3e7c23fe3384beb18f18d7749f2b4a"
1302
+ },
1303
+ {
1304
+ "demo_id": "demo_12",
1305
+ "file": "demo_12.h5",
1306
+ "length": 485,
1307
+ "fields": [
1308
+ {
1309
+ "path": "actions/panda_leaphand-1",
1310
+ "shape": [
1311
+ 485,
1312
+ 22
1313
+ ],
1314
+ "dtype": "float32"
1315
+ },
1316
+ {
1317
+ "path": "actions/panda_leaphand_left-0",
1318
+ "shape": [
1319
+ 485,
1320
+ 22
1321
+ ],
1322
+ "dtype": "float32"
1323
+ },
1324
+ {
1325
+ "path": "obs/agent/panda_leaphand-1/qpos",
1326
+ "shape": [
1327
+ 486,
1328
+ 23
1329
+ ],
1330
+ "dtype": "float32"
1331
+ },
1332
+ {
1333
+ "path": "obs/agent/panda_leaphand-1/qvel",
1334
+ "shape": [
1335
+ 486,
1336
+ 23
1337
+ ],
1338
+ "dtype": "float32"
1339
+ },
1340
+ {
1341
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
1342
+ "shape": [
1343
+ 486,
1344
+ 23
1345
+ ],
1346
+ "dtype": "float32"
1347
+ },
1348
+ {
1349
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
1350
+ "shape": [
1351
+ 486,
1352
+ 23
1353
+ ],
1354
+ "dtype": "float32"
1355
+ },
1356
+ {
1357
+ "path": "obs/sensor_data/front_camera/rgb",
1358
+ "shape": [
1359
+ 486,
1360
+ 240,
1361
+ 320,
1362
+ 3
1363
+ ],
1364
+ "dtype": "uint8"
1365
+ },
1366
+ {
1367
+ "path": "obs/sensor_data/left_camera/rgb",
1368
+ "shape": [
1369
+ 486,
1370
+ 240,
1371
+ 320,
1372
+ 3
1373
+ ],
1374
+ "dtype": "uint8"
1375
+ },
1376
+ {
1377
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1378
+ "shape": [
1379
+ 486,
1380
+ 240,
1381
+ 320,
1382
+ 3
1383
+ ],
1384
+ "dtype": "uint8"
1385
+ },
1386
+ {
1387
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1388
+ "shape": [
1389
+ 486,
1390
+ 240,
1391
+ 320,
1392
+ 3
1393
+ ],
1394
+ "dtype": "uint8"
1395
+ },
1396
+ {
1397
+ "path": "obs/sensor_data/top_camera/rgb",
1398
+ "shape": [
1399
+ 486,
1400
+ 320,
1401
+ 240,
1402
+ 3
1403
+ ],
1404
+ "dtype": "uint8"
1405
+ }
1406
+ ],
1407
+ "trajectory_sha256": "9547de4c2c5bbda2d520ce7dcb1b3ada3aac5bcd40e7edf873bd565778cf72a9",
1408
+ "source_sha256": "498e69f64927d672232e8de8f51155280c2ce476cf6fbf60a544877cfd506280",
1409
+ "view_sha256": "15dc0edc5f808474a0169e47048e3863a63439b75d5fc97a14a3d77e6732a967"
1410
+ },
1411
+ {
1412
+ "demo_id": "demo_13",
1413
+ "file": "demo_13.h5",
1414
+ "length": 479,
1415
+ "fields": [
1416
+ {
1417
+ "path": "actions/panda_leaphand-1",
1418
+ "shape": [
1419
+ 479,
1420
+ 22
1421
+ ],
1422
+ "dtype": "float32"
1423
+ },
1424
+ {
1425
+ "path": "actions/panda_leaphand_left-0",
1426
+ "shape": [
1427
+ 479,
1428
+ 22
1429
+ ],
1430
+ "dtype": "float32"
1431
+ },
1432
+ {
1433
+ "path": "obs/agent/panda_leaphand-1/qpos",
1434
+ "shape": [
1435
+ 480,
1436
+ 23
1437
+ ],
1438
+ "dtype": "float32"
1439
+ },
1440
+ {
1441
+ "path": "obs/agent/panda_leaphand-1/qvel",
1442
+ "shape": [
1443
+ 480,
1444
+ 23
1445
+ ],
1446
+ "dtype": "float32"
1447
+ },
1448
+ {
1449
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
1450
+ "shape": [
1451
+ 480,
1452
+ 23
1453
+ ],
1454
+ "dtype": "float32"
1455
+ },
1456
+ {
1457
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
1458
+ "shape": [
1459
+ 480,
1460
+ 23
1461
+ ],
1462
+ "dtype": "float32"
1463
+ },
1464
+ {
1465
+ "path": "obs/sensor_data/front_camera/rgb",
1466
+ "shape": [
1467
+ 480,
1468
+ 240,
1469
+ 320,
1470
+ 3
1471
+ ],
1472
+ "dtype": "uint8"
1473
+ },
1474
+ {
1475
+ "path": "obs/sensor_data/left_camera/rgb",
1476
+ "shape": [
1477
+ 480,
1478
+ 240,
1479
+ 320,
1480
+ 3
1481
+ ],
1482
+ "dtype": "uint8"
1483
+ },
1484
+ {
1485
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1486
+ "shape": [
1487
+ 480,
1488
+ 240,
1489
+ 320,
1490
+ 3
1491
+ ],
1492
+ "dtype": "uint8"
1493
+ },
1494
+ {
1495
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1496
+ "shape": [
1497
+ 480,
1498
+ 240,
1499
+ 320,
1500
+ 3
1501
+ ],
1502
+ "dtype": "uint8"
1503
+ },
1504
+ {
1505
+ "path": "obs/sensor_data/top_camera/rgb",
1506
+ "shape": [
1507
+ 480,
1508
+ 320,
1509
+ 240,
1510
+ 3
1511
+ ],
1512
+ "dtype": "uint8"
1513
+ }
1514
+ ],
1515
+ "trajectory_sha256": "968bebc699f9f3a6678b9d60095d947eb33acda5a22fb467f2e847a7da8f5a9c",
1516
+ "source_sha256": "fd68b2805589fe1bd9faee6214a8020901ab643b5bc44327ecd4d74f38a08a4e",
1517
+ "view_sha256": "44187d6446b3d77edc706f2fed757d24ea644c6fb251824ad033d75b7b012f8b"
1518
+ },
1519
+ {
1520
+ "demo_id": "demo_14",
1521
+ "file": "demo_14.h5",
1522
+ "length": 484,
1523
+ "fields": [
1524
+ {
1525
+ "path": "actions/panda_leaphand-1",
1526
+ "shape": [
1527
+ 484,
1528
+ 22
1529
+ ],
1530
+ "dtype": "float32"
1531
+ },
1532
+ {
1533
+ "path": "actions/panda_leaphand_left-0",
1534
+ "shape": [
1535
+ 484,
1536
+ 22
1537
+ ],
1538
+ "dtype": "float32"
1539
+ },
1540
+ {
1541
+ "path": "obs/agent/panda_leaphand-1/qpos",
1542
+ "shape": [
1543
+ 485,
1544
+ 23
1545
+ ],
1546
+ "dtype": "float32"
1547
+ },
1548
+ {
1549
+ "path": "obs/agent/panda_leaphand-1/qvel",
1550
+ "shape": [
1551
+ 485,
1552
+ 23
1553
+ ],
1554
+ "dtype": "float32"
1555
+ },
1556
+ {
1557
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
1558
+ "shape": [
1559
+ 485,
1560
+ 23
1561
+ ],
1562
+ "dtype": "float32"
1563
+ },
1564
+ {
1565
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
1566
+ "shape": [
1567
+ 485,
1568
+ 23
1569
+ ],
1570
+ "dtype": "float32"
1571
+ },
1572
+ {
1573
+ "path": "obs/sensor_data/front_camera/rgb",
1574
+ "shape": [
1575
+ 485,
1576
+ 240,
1577
+ 320,
1578
+ 3
1579
+ ],
1580
+ "dtype": "uint8"
1581
+ },
1582
+ {
1583
+ "path": "obs/sensor_data/left_camera/rgb",
1584
+ "shape": [
1585
+ 485,
1586
+ 240,
1587
+ 320,
1588
+ 3
1589
+ ],
1590
+ "dtype": "uint8"
1591
+ },
1592
+ {
1593
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1594
+ "shape": [
1595
+ 485,
1596
+ 240,
1597
+ 320,
1598
+ 3
1599
+ ],
1600
+ "dtype": "uint8"
1601
+ },
1602
+ {
1603
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1604
+ "shape": [
1605
+ 485,
1606
+ 240,
1607
+ 320,
1608
+ 3
1609
+ ],
1610
+ "dtype": "uint8"
1611
+ },
1612
+ {
1613
+ "path": "obs/sensor_data/top_camera/rgb",
1614
+ "shape": [
1615
+ 485,
1616
+ 320,
1617
+ 240,
1618
+ 3
1619
+ ],
1620
+ "dtype": "uint8"
1621
+ }
1622
+ ],
1623
+ "trajectory_sha256": "a324c753e9785c20bb351355f589039e81f2c7edc534292c32e12f45d4c3515d",
1624
+ "source_sha256": "ce2a96aca63392119c874b2eb669c7d4b8f04410213a384f15c89714b0ecb550",
1625
+ "view_sha256": "b7ebcc1f4bc491c919e87bb001943d19c112d8418147e565f304762f2094d07c"
1626
+ },
1627
+ {
1628
+ "demo_id": "demo_15",
1629
+ "file": "demo_15.h5",
1630
+ "length": 485,
1631
+ "fields": [
1632
+ {
1633
+ "path": "actions/panda_leaphand-1",
1634
+ "shape": [
1635
+ 485,
1636
+ 22
1637
+ ],
1638
+ "dtype": "float32"
1639
+ },
1640
+ {
1641
+ "path": "actions/panda_leaphand_left-0",
1642
+ "shape": [
1643
+ 485,
1644
+ 22
1645
+ ],
1646
+ "dtype": "float32"
1647
+ },
1648
+ {
1649
+ "path": "obs/agent/panda_leaphand-1/qpos",
1650
+ "shape": [
1651
+ 486,
1652
+ 23
1653
+ ],
1654
+ "dtype": "float32"
1655
+ },
1656
+ {
1657
+ "path": "obs/agent/panda_leaphand-1/qvel",
1658
+ "shape": [
1659
+ 486,
1660
+ 23
1661
+ ],
1662
+ "dtype": "float32"
1663
+ },
1664
+ {
1665
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
1666
+ "shape": [
1667
+ 486,
1668
+ 23
1669
+ ],
1670
+ "dtype": "float32"
1671
+ },
1672
+ {
1673
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
1674
+ "shape": [
1675
+ 486,
1676
+ 23
1677
+ ],
1678
+ "dtype": "float32"
1679
+ },
1680
+ {
1681
+ "path": "obs/sensor_data/front_camera/rgb",
1682
+ "shape": [
1683
+ 486,
1684
+ 240,
1685
+ 320,
1686
+ 3
1687
+ ],
1688
+ "dtype": "uint8"
1689
+ },
1690
+ {
1691
+ "path": "obs/sensor_data/left_camera/rgb",
1692
+ "shape": [
1693
+ 486,
1694
+ 240,
1695
+ 320,
1696
+ 3
1697
+ ],
1698
+ "dtype": "uint8"
1699
+ },
1700
+ {
1701
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1702
+ "shape": [
1703
+ 486,
1704
+ 240,
1705
+ 320,
1706
+ 3
1707
+ ],
1708
+ "dtype": "uint8"
1709
+ },
1710
+ {
1711
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1712
+ "shape": [
1713
+ 486,
1714
+ 240,
1715
+ 320,
1716
+ 3
1717
+ ],
1718
+ "dtype": "uint8"
1719
+ },
1720
+ {
1721
+ "path": "obs/sensor_data/top_camera/rgb",
1722
+ "shape": [
1723
+ 486,
1724
+ 320,
1725
+ 240,
1726
+ 3
1727
+ ],
1728
+ "dtype": "uint8"
1729
+ }
1730
+ ],
1731
+ "trajectory_sha256": "5643d89c5679b954e723b956aec2c77e2cf6ea135d30a53406d5a4b1e2fc12dc",
1732
+ "source_sha256": "3017eb270c103ce55ed2b8c6600b08114ef0fb597a00f536a214e041b6080c2b",
1733
+ "view_sha256": "100b37f92e1eb56c1dbc1e66f5bc6c07a3b82c652aafaa10bcf7b3fb195533b5"
1734
+ },
1735
+ {
1736
+ "demo_id": "demo_16",
1737
+ "file": "demo_16.h5",
1738
+ "length": 482,
1739
+ "fields": [
1740
+ {
1741
+ "path": "actions/panda_leaphand-1",
1742
+ "shape": [
1743
+ 482,
1744
+ 22
1745
+ ],
1746
+ "dtype": "float32"
1747
+ },
1748
+ {
1749
+ "path": "actions/panda_leaphand_left-0",
1750
+ "shape": [
1751
+ 482,
1752
+ 22
1753
+ ],
1754
+ "dtype": "float32"
1755
+ },
1756
+ {
1757
+ "path": "obs/agent/panda_leaphand-1/qpos",
1758
+ "shape": [
1759
+ 483,
1760
+ 23
1761
+ ],
1762
+ "dtype": "float32"
1763
+ },
1764
+ {
1765
+ "path": "obs/agent/panda_leaphand-1/qvel",
1766
+ "shape": [
1767
+ 483,
1768
+ 23
1769
+ ],
1770
+ "dtype": "float32"
1771
+ },
1772
+ {
1773
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
1774
+ "shape": [
1775
+ 483,
1776
+ 23
1777
+ ],
1778
+ "dtype": "float32"
1779
+ },
1780
+ {
1781
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
1782
+ "shape": [
1783
+ 483,
1784
+ 23
1785
+ ],
1786
+ "dtype": "float32"
1787
+ },
1788
+ {
1789
+ "path": "obs/sensor_data/front_camera/rgb",
1790
+ "shape": [
1791
+ 483,
1792
+ 240,
1793
+ 320,
1794
+ 3
1795
+ ],
1796
+ "dtype": "uint8"
1797
+ },
1798
+ {
1799
+ "path": "obs/sensor_data/left_camera/rgb",
1800
+ "shape": [
1801
+ 483,
1802
+ 240,
1803
+ 320,
1804
+ 3
1805
+ ],
1806
+ "dtype": "uint8"
1807
+ },
1808
+ {
1809
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1810
+ "shape": [
1811
+ 483,
1812
+ 240,
1813
+ 320,
1814
+ 3
1815
+ ],
1816
+ "dtype": "uint8"
1817
+ },
1818
+ {
1819
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1820
+ "shape": [
1821
+ 483,
1822
+ 240,
1823
+ 320,
1824
+ 3
1825
+ ],
1826
+ "dtype": "uint8"
1827
+ },
1828
+ {
1829
+ "path": "obs/sensor_data/top_camera/rgb",
1830
+ "shape": [
1831
+ 483,
1832
+ 320,
1833
+ 240,
1834
+ 3
1835
+ ],
1836
+ "dtype": "uint8"
1837
+ }
1838
+ ],
1839
+ "trajectory_sha256": "ea318c5f8387109babe622c73fcf4c9553644e7a18d4ca45d64ae5b8e82a4234",
1840
+ "source_sha256": "d671b9acedab2f63ff370a8c8b26af2a2b8649501e51f12d858b9152fbcecd8d",
1841
+ "view_sha256": "83aedc0863227d57549ec16a9eb1fa5eed3b0af0e1da02e4faf7b80ac575fecb"
1842
+ },
1843
+ {
1844
+ "demo_id": "demo_17",
1845
+ "file": "demo_17.h5",
1846
+ "length": 484,
1847
+ "fields": [
1848
+ {
1849
+ "path": "actions/panda_leaphand-1",
1850
+ "shape": [
1851
+ 484,
1852
+ 22
1853
+ ],
1854
+ "dtype": "float32"
1855
+ },
1856
+ {
1857
+ "path": "actions/panda_leaphand_left-0",
1858
+ "shape": [
1859
+ 484,
1860
+ 22
1861
+ ],
1862
+ "dtype": "float32"
1863
+ },
1864
+ {
1865
+ "path": "obs/agent/panda_leaphand-1/qpos",
1866
+ "shape": [
1867
+ 485,
1868
+ 23
1869
+ ],
1870
+ "dtype": "float32"
1871
+ },
1872
+ {
1873
+ "path": "obs/agent/panda_leaphand-1/qvel",
1874
+ "shape": [
1875
+ 485,
1876
+ 23
1877
+ ],
1878
+ "dtype": "float32"
1879
+ },
1880
+ {
1881
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
1882
+ "shape": [
1883
+ 485,
1884
+ 23
1885
+ ],
1886
+ "dtype": "float32"
1887
+ },
1888
+ {
1889
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
1890
+ "shape": [
1891
+ 485,
1892
+ 23
1893
+ ],
1894
+ "dtype": "float32"
1895
+ },
1896
+ {
1897
+ "path": "obs/sensor_data/front_camera/rgb",
1898
+ "shape": [
1899
+ 485,
1900
+ 240,
1901
+ 320,
1902
+ 3
1903
+ ],
1904
+ "dtype": "uint8"
1905
+ },
1906
+ {
1907
+ "path": "obs/sensor_data/left_camera/rgb",
1908
+ "shape": [
1909
+ 485,
1910
+ 240,
1911
+ 320,
1912
+ 3
1913
+ ],
1914
+ "dtype": "uint8"
1915
+ },
1916
+ {
1917
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1918
+ "shape": [
1919
+ 485,
1920
+ 240,
1921
+ 320,
1922
+ 3
1923
+ ],
1924
+ "dtype": "uint8"
1925
+ },
1926
+ {
1927
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1928
+ "shape": [
1929
+ 485,
1930
+ 240,
1931
+ 320,
1932
+ 3
1933
+ ],
1934
+ "dtype": "uint8"
1935
+ },
1936
+ {
1937
+ "path": "obs/sensor_data/top_camera/rgb",
1938
+ "shape": [
1939
+ 485,
1940
+ 320,
1941
+ 240,
1942
+ 3
1943
+ ],
1944
+ "dtype": "uint8"
1945
+ }
1946
+ ],
1947
+ "trajectory_sha256": "50270a02e68e9b0679c5a5e3fbe73fdad92a3a10f1725962d37b1200f973dc00",
1948
+ "source_sha256": "f5ec32cefc43c9c56013789dafc33549ba06089a1c01eda2ba8c44efd97dc1f6",
1949
+ "view_sha256": "b771d3b16928865cf6d1d5ddf172d1521aa039f729544211d9616e34eb244bc0"
1950
+ },
1951
+ {
1952
+ "demo_id": "demo_18",
1953
+ "file": "demo_18.h5",
1954
+ "length": 484,
1955
+ "fields": [
1956
+ {
1957
+ "path": "actions/panda_leaphand-1",
1958
+ "shape": [
1959
+ 484,
1960
+ 22
1961
+ ],
1962
+ "dtype": "float32"
1963
+ },
1964
+ {
1965
+ "path": "actions/panda_leaphand_left-0",
1966
+ "shape": [
1967
+ 484,
1968
+ 22
1969
+ ],
1970
+ "dtype": "float32"
1971
+ },
1972
+ {
1973
+ "path": "obs/agent/panda_leaphand-1/qpos",
1974
+ "shape": [
1975
+ 485,
1976
+ 23
1977
+ ],
1978
+ "dtype": "float32"
1979
+ },
1980
+ {
1981
+ "path": "obs/agent/panda_leaphand-1/qvel",
1982
+ "shape": [
1983
+ 485,
1984
+ 23
1985
+ ],
1986
+ "dtype": "float32"
1987
+ },
1988
+ {
1989
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
1990
+ "shape": [
1991
+ 485,
1992
+ 23
1993
+ ],
1994
+ "dtype": "float32"
1995
+ },
1996
+ {
1997
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
1998
+ "shape": [
1999
+ 485,
2000
+ 23
2001
+ ],
2002
+ "dtype": "float32"
2003
+ },
2004
+ {
2005
+ "path": "obs/sensor_data/front_camera/rgb",
2006
+ "shape": [
2007
+ 485,
2008
+ 240,
2009
+ 320,
2010
+ 3
2011
+ ],
2012
+ "dtype": "uint8"
2013
+ },
2014
+ {
2015
+ "path": "obs/sensor_data/left_camera/rgb",
2016
+ "shape": [
2017
+ 485,
2018
+ 240,
2019
+ 320,
2020
+ 3
2021
+ ],
2022
+ "dtype": "uint8"
2023
+ },
2024
+ {
2025
+ "path": "obs/sensor_data/robot_camera_0/rgb",
2026
+ "shape": [
2027
+ 485,
2028
+ 240,
2029
+ 320,
2030
+ 3
2031
+ ],
2032
+ "dtype": "uint8"
2033
+ },
2034
+ {
2035
+ "path": "obs/sensor_data/robot_camera_1/rgb",
2036
+ "shape": [
2037
+ 485,
2038
+ 240,
2039
+ 320,
2040
+ 3
2041
+ ],
2042
+ "dtype": "uint8"
2043
+ },
2044
+ {
2045
+ "path": "obs/sensor_data/top_camera/rgb",
2046
+ "shape": [
2047
+ 485,
2048
+ 320,
2049
+ 240,
2050
+ 3
2051
+ ],
2052
+ "dtype": "uint8"
2053
+ }
2054
+ ],
2055
+ "trajectory_sha256": "73fbf52b3659a41d553908c6a9b506840f0ac78027893e010ca70c2e9787f6ab",
2056
+ "source_sha256": "dc8792bf7222e329dbabcbd57d17fc8bfc4400e72a87253a17bf681dda28c106",
2057
+ "view_sha256": "6104efa473760e7d71c9ed2e53d3f588f1998c95f40248bc8d0bfc556d15ebfb"
2058
+ },
2059
+ {
2060
+ "demo_id": "demo_19",
2061
+ "file": "demo_19.h5",
2062
+ "length": 483,
2063
+ "fields": [
2064
+ {
2065
+ "path": "actions/panda_leaphand-1",
2066
+ "shape": [
2067
+ 483,
2068
+ 22
2069
+ ],
2070
+ "dtype": "float32"
2071
+ },
2072
+ {
2073
+ "path": "actions/panda_leaphand_left-0",
2074
+ "shape": [
2075
+ 483,
2076
+ 22
2077
+ ],
2078
+ "dtype": "float32"
2079
+ },
2080
+ {
2081
+ "path": "obs/agent/panda_leaphand-1/qpos",
2082
+ "shape": [
2083
+ 484,
2084
+ 23
2085
+ ],
2086
+ "dtype": "float32"
2087
+ },
2088
+ {
2089
+ "path": "obs/agent/panda_leaphand-1/qvel",
2090
+ "shape": [
2091
+ 484,
2092
+ 23
2093
+ ],
2094
+ "dtype": "float32"
2095
+ },
2096
+ {
2097
+ "path": "obs/agent/panda_leaphand_left-0/qpos",
2098
+ "shape": [
2099
+ 484,
2100
+ 23
2101
+ ],
2102
+ "dtype": "float32"
2103
+ },
2104
+ {
2105
+ "path": "obs/agent/panda_leaphand_left-0/qvel",
2106
+ "shape": [
2107
+ 484,
2108
+ 23
2109
+ ],
2110
+ "dtype": "float32"
2111
+ },
2112
+ {
2113
+ "path": "obs/sensor_data/front_camera/rgb",
2114
+ "shape": [
2115
+ 484,
2116
+ 240,
2117
+ 320,
2118
+ 3
2119
+ ],
2120
+ "dtype": "uint8"
2121
+ },
2122
+ {
2123
+ "path": "obs/sensor_data/left_camera/rgb",
2124
+ "shape": [
2125
+ 484,
2126
+ 240,
2127
+ 320,
2128
+ 3
2129
+ ],
2130
+ "dtype": "uint8"
2131
+ },
2132
+ {
2133
+ "path": "obs/sensor_data/robot_camera_0/rgb",
2134
+ "shape": [
2135
+ 484,
2136
+ 240,
2137
+ 320,
2138
+ 3
2139
+ ],
2140
+ "dtype": "uint8"
2141
+ },
2142
+ {
2143
+ "path": "obs/sensor_data/robot_camera_1/rgb",
2144
+ "shape": [
2145
+ 484,
2146
+ 240,
2147
+ 320,
2148
+ 3
2149
+ ],
2150
+ "dtype": "uint8"
2151
+ },
2152
+ {
2153
+ "path": "obs/sensor_data/top_camera/rgb",
2154
+ "shape": [
2155
+ 484,
2156
+ 320,
2157
+ 240,
2158
+ 3
2159
+ ],
2160
+ "dtype": "uint8"
2161
+ }
2162
+ ],
2163
+ "trajectory_sha256": "d58930c7326c9bdc7a0b402c7de8c64ebc3a370d1e28347a60697dd08de28471",
2164
+ "source_sha256": "ddcb0a9f4db52f5fa9e6e43dba8ab97a1766dbff4bc08b0df0f6005c78dc1efc",
2165
+ "view_sha256": "222aae248b23fc50e0180914fa559c65255776af1b030c506245cb730e737931"
2166
+ }
2167
+ ]
2168
+ }
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/environment.lock ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ absl-py==2.4.0
2
+ accelerate==1.13.0
3
+ addict==2.4.0
4
+ aiohappyeyeballs==2.6.1
5
+ aiohttp==3.13.5
6
+ aiosignal==1.4.0
7
+ annotated-types==0.7.0
8
+ anyio==4.13.0
9
+ anytree==2.13.0
10
+ asttokens==3.0.1
11
+ async-timeout==5.0.1
12
+ attrs==26.1.0
13
+ av==15.1.0
14
+ blinker==1.9.0
15
+ casadi==3.7.2
16
+ certifi==2026.4.22
17
+ cffi==2.0.0
18
+ charset-normalizer==3.4.7
19
+ click==8.3.3
20
+ cloudpickle==3.1.2
21
+ cmake==4.1.3
22
+ coacd==1.0.11
23
+ colorama==0.4.6
24
+ colorlog==6.10.1
25
+ comm==0.2.3
26
+ ConfigArgParse==1.7.5
27
+ contourpy==1.3.2
28
+ cuda-bindings==12.9.4
29
+ cuda-pathfinder==1.5.4
30
+ cycler==0.12.1
31
+ dacite==1.9.2
32
+ dash==4.1.0
33
+ dataclasses-json==0.6.7
34
+ datasets==4.8.5
35
+ decorator==5.2.1
36
+ deepdiff==8.6.2
37
+ Deprecated==1.3.1
38
+ diffusers==0.35.2
39
+ dill==0.4.1
40
+ docstring_parser==0.18.0
41
+ draccus==0.10.0
42
+ einops==0.8.2
43
+ embreex==4.4.0
44
+ etils==1.13.0
45
+ evdev==1.9.3
46
+ exceptiongroup==1.3.1
47
+ executing==2.2.1
48
+ Farama-Notifications==0.0.6
49
+ fast_simplification==0.1.13
50
+ fastjsonschema==2.21.2
51
+ filelock==3.29.0
52
+ Flask==3.1.3
53
+ fonttools==4.63.0
54
+ freetype-py==2.5.1
55
+ frozenlist==1.8.0
56
+ fsspec==2026.2.0
57
+ genesis-world @ file:///data/gck/genesis-world
58
+ git-filter-repo==2.47.0
59
+ gitdb==4.0.12
60
+ GitPython==3.1.50
61
+ glfw==2.10.0
62
+ gs-madrona==0.0.5
63
+ gstaichi==3.3.0
64
+ gym==0.26.2
65
+ gym-notices==0.1.0
66
+ gymnasium==1.3.0
67
+ h11==0.16.0
68
+ h5py==3.16.0
69
+ hf-xet==1.5.0
70
+ hf_transfer==0.1.9
71
+ httpcore==1.0.9
72
+ httpx==0.28.1
73
+ huggingface-hub==0.35.3
74
+ idna==3.15
75
+ ImageIO==2.37.3
76
+ imageio-ffmpeg==0.6.0
77
+ importlib_metadata==9.0.0
78
+ importlib_resources==7.1.0
79
+ inquirerpy==0.3.4
80
+ ipython==8.39.0
81
+ ipywidgets==8.1.8
82
+ itsdangerous==2.2.0
83
+ jedi==0.20.0
84
+ Jinja2==3.1.6
85
+ joblib==1.5.3
86
+ jsonlines==4.0.0
87
+ jsonschema==4.26.0
88
+ jsonschema-specifications==2025.9.1
89
+ jupyter_core==5.9.1
90
+ jupyterlab_widgets==3.0.16
91
+ kiwisolver==1.5.0
92
+ lazy-loader==0.5
93
+ lerobot==0.4.4
94
+ libigl==2.6.2
95
+ llvmlite==0.47.0
96
+ loguru==0.7.3
97
+ lxml==6.1.0
98
+ manifold3d==3.4.1
99
+ -e git+ssh://git@github.com/HeegerGao/maniladder.git@f6eed17496816bb41ce484f2d7580990a908d3e6#egg=maniladder
100
+ mapbox_earcut==2.0.0
101
+ markdown-it-py==4.2.0
102
+ MarkupSafe==3.0.3
103
+ marshmallow==3.26.2
104
+ matplotlib==3.10.9
105
+ matplotlib-inline==0.2.2
106
+ mdurl==0.1.2
107
+ mergedeep==1.3.4
108
+ moviepy==2.2.1
109
+ mpmath==1.3.0
110
+ msgspec==0.21.1
111
+ mujoco==3.8.1
112
+ multidict==6.7.1
113
+ multiprocess==0.70.19
114
+ mypy_extensions==1.1.0
115
+ nbformat==5.10.4
116
+ nest-asyncio==1.6.0
117
+ networkx==3.4.2
118
+ ninja==1.13.0
119
+ nlopt==2.10.0
120
+ numba==0.65.1
121
+ numpy @ file:///home/conda/feedstock_root/build_artifacts/numpy_1747544640217/work/dist/numpy-2.2.6-cp310-cp310-linux_x86_64.whl#sha256=d6d964caeef85d00073d27cd62b46883d275b3d8162f723f0fcabbd0b3cc3f9d
122
+ nvidia-cublas-cu12==12.8.4.1
123
+ nvidia-cuda-cupti-cu12==12.8.90
124
+ nvidia-cuda-nvrtc-cu12==12.8.93
125
+ nvidia-cuda-runtime-cu12==12.8.90
126
+ nvidia-cudnn-cu12==9.10.2.21
127
+ nvidia-cufft-cu12==11.3.3.83
128
+ nvidia-cufile-cu12==1.13.1.3
129
+ nvidia-curand-cu12==10.3.9.90
130
+ nvidia-cusolver-cu12==11.7.3.90
131
+ nvidia-cusparse-cu12==12.5.8.93
132
+ nvidia-cusparselt-cu12==0.7.1
133
+ nvidia-nccl-cu12==2.27.5
134
+ nvidia-nvjitlink-cu12==12.8.93
135
+ nvidia-nvshmem-cu12==3.4.5
136
+ nvidia-nvtx-cu12==12.8.90
137
+ open3d==0.19.0
138
+ opencv-python==4.13.0.92
139
+ opencv-python-headless==4.12.0.88
140
+ OpenEXR==3.4.11
141
+ orderly-set==5.5.0
142
+ packaging==25.0
143
+ pandas==2.3.3
144
+ parso==0.8.7
145
+ pexpect==4.9.0
146
+ pfzy==0.3.4
147
+ pillow==11.3.0
148
+ pin==4.0.0
149
+ pip @ file:///home/task_177498883471825/croot/pip_1774988882575/work
150
+ platformdirs==4.9.6
151
+ plotly==5.24.1
152
+ polyscope==2.6.1
153
+ proglog==0.1.12
154
+ prompt_toolkit==3.0.52
155
+ propcache==0.5.2
156
+ protobuf==6.33.6
157
+ psutil==7.2.2
158
+ ptyprocess==0.7.0
159
+ pure_eval==0.2.3
160
+ py-cpuinfo==9.0.0
161
+ py-spy==0.4.2
162
+ pyarrow==24.0.0
163
+ pycollada==0.9.3
164
+ pycparser==3.0
165
+ pydantic==2.13.4
166
+ pydantic_core==2.46.4
167
+ PyGEL3D==0.6.1
168
+ pyglet==2.1.14
169
+ pygltflib==1.16.0
170
+ Pygments==2.20.0
171
+ pymeshlab==2025.7.post1
172
+ pynput==1.8.2
173
+ PyOpenGL==3.1.10
174
+ pyparsing==3.3.2
175
+ pyperclip==1.11.0
176
+ PyQt6==6.11.0
177
+ PyQt6-Qt6==6.11.0
178
+ PyQt6_sip==13.11.1
179
+ pyquaternion==0.9.9
180
+ pyserial==3.5
181
+ pysplashsurf==0.12.0.2
182
+ python-dateutil==2.9.0.post0
183
+ python-dotenv==1.2.2
184
+ python-xlib==0.33
185
+ pytransform3d==3.15.0
186
+ pytz==2026.2
187
+ pyuipc @ file:///tmp/claude-1005/-mnt4-gck-maniladder/9f005e67-04f6-4368-bd0e-9f40896c7dbf/scratchpad/pyuipc_dl/pyuipc-0.0.25-cp310-cp310-manylinux_2_34_x86_64.whl#sha256=76ff13554e1971ea4401db2c875369cf32352ae3a5a4c398c45193231b5cbf00
188
+ PyYAML==6.0.3
189
+ pyyaml-include==1.4.1
190
+ referencing==0.37.0
191
+ regex==2026.5.9
192
+ requests==2.34.2
193
+ rerun-sdk==0.26.2
194
+ retrying==1.4.2
195
+ rich==14.3.4
196
+ rpds-py==0.30.0
197
+ rtree==1.4.1
198
+ safetensors==0.7.0
199
+ sapien==3.0.3
200
+ scikit-image==0.25.2
201
+ scikit-learn==1.7.2
202
+ scipy @ file:///home/conda/feedstock_root/build_artifacts/scipy-split_1739790642651/work/dist/scipy-1.15.2-cp310-cp310-linux_x86_64.whl#sha256=9e52bad6c3294d1a5b04a13632118ca2157130603c6c018c2d710162b223b27e
203
+ sentry-sdk==2.60.0
204
+ setuptools==80.10.2
205
+ shapely==2.1.2
206
+ six==1.17.0
207
+ smmap==5.0.3
208
+ stack-data==0.6.3
209
+ svg.path==7.0
210
+ sympy==1.14.0
211
+ tenacity==9.1.4
212
+ termcolor==3.3.0
213
+ tetgen==0.8.4
214
+ threadpoolctl==3.6.0
215
+ tifffile==2025.5.10
216
+ timm==1.0.27
217
+ toml==0.10.2
218
+ torch==2.10.0
219
+ torchaudio==2.11.0
220
+ torchcodec==0.10.0
221
+ torchvision==0.25.0
222
+ tqdm==4.67.3
223
+ traitlets==5.15.0
224
+ transforms3d==0.4.2
225
+ trimesh==4.12.2
226
+ triton==3.6.0
227
+ typeguard==4.5.2
228
+ typing-inspect==0.9.0
229
+ typing-inspection==0.4.2
230
+ typing_extensions==4.15.0
231
+ tyro==1.0.13
232
+ tzdata==2026.2
233
+ urllib3==2.7.0
234
+ vhacdx==0.0.10
235
+ viser==1.0.27
236
+ vtk==9.6.1
237
+ wandb==0.24.2
238
+ wcwidth==0.7.0
239
+ websockets==16.0
240
+ Werkzeug==3.1.8
241
+ wheel==0.46.3
242
+ widgetsnbextension==4.0.15
243
+ wrapt==2.1.2
244
+ xxhash==3.7.0
245
+ yarl==1.23.0
246
+ yourdfpy==0.0.60
247
+ z3-solver==4.16.0.0
248
+ zipp==3.23.1
249
+ zstandard==0.25.0
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/features.py ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Handwritten RGB features; no pretrained or fitted vision weights."""
2
+ import cv2
3
+ import numpy as np
4
+ CAMERAS=('front_camera','top_camera')
5
+ PATHS=('agent/panda_leaphand-1/qpos','agent/panda_leaphand-1/qvel','agent/panda_leaphand_left-0/qpos','agent/panda_leaphand_left-0/qvel')
6
+ def visual(rgb):
7
+ out=[]
8
+ for cam in CAMERAS:
9
+ im=np.asarray(rgb[cam]);im=im[0] if im.ndim==4 and im.shape[0]==1 else im
10
+ if im.ndim!=3 or im.shape[2]!=3: raise ValueError('Expected RGB HWC or 1HWC')
11
+ hsv=cv2.cvtColor(im.astype(np.uint8),cv2.COLOR_RGB2HSV)
12
+ h,s,v=cv2.split(hsv)
13
+ for lo,hi in ((35,90),(95,135),(0,12),(165,179)):
14
+ mask=((h>=lo)&(h<=hi)&(s>100)&(v>60)).astype(np.uint8)
15
+ count,labels,stats,cent=cv2.connectedComponentsWithStats(mask)
16
+ if count>1 and stats[1:,4].max()>=3:
17
+ k=1+np.argmax(stats[1:,4]);x,y=cent[k];area=stats[k,4]
18
+ out.extend([x/im.shape[1],y/im.shape[0],np.sqrt(area/(im.shape[0]*im.shape[1])),1.])
19
+ else: out.extend([0.,0.,0.,0.])
20
+ out.extend((cv2.resize(im,(16,12),interpolation=cv2.INTER_AREA).astype(np.float32)/255).ravel())
21
+ return np.asarray(out,np.float32)
22
+ def proprio(p):
23
+ out=[]
24
+ for path in PATHS:
25
+ x=np.asarray(p[path],np.float32).reshape(-1)
26
+ if x.size!=23 or not np.isfinite(x).all(): raise ValueError('Invalid proprioception: '+path)
27
+ out.extend(x)
28
+ return np.array(out,np.float32)
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/findings.md ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Live E20 exploration
2
+
3
+ Budget initially 0/20; API budget_status was first successful environment call. Initial two local import/invocation errors did not call service. Contract and protocol read. Only task-local saved D20 compilation used.
4
+
5
+ Public visible rule preregistered: true only if legal camera images unambiguously show three balls participating in repeated coordinated catch/throw cycles. Two held balls, occlusion, or ambiguity are false. Sparse native success is recorded separately when supplied.
6
+
7
+ Episode 1: initial harness failed after first action because execute_actions wraps returns in steps; fixed wrapper, resumed via latest_observation without resetting/replaying. That first action was baseline controller action at initial observation; subsequent actions saved in ep01/actions.npy beginning native step 2. Observation step_id is outside payload; controller now also maintains fallback call counter for raw observations. First 150 steps track demonstration proprioception closely. Front at 100 shows ball contact; front at 200 shows both colored balls raised. No third ball established. Full observations include all legal cameras/proprioception every 50 steps.
8
+
9
+ Episode 1 terminal API observation at native step 456: terminated=true, success=true. This is native sparse success, distinct from public visible rule (false: no three-ball continuous cycle established). Old harness failed to inspect nested termination and attempted one rejected action after terminal; no extra step/episode charged. budget reports one slot and 456 steps. finish_episode acknowledged. Live observation and images preserved at ep01/obs_456.npz. Baseline action gain was 1.
10
+
11
+ Episode 2 experiment: gain 2 applied to retrieved normalized action with contract clipping, maximum observation-controlled duration 300; tests timing robustness and whether faster movement preserves grasps. No trajectory replay.
12
+
13
+ Episode 2 completed 300 steps, native success false. Gain 2 reached demo-like grasp/lift in roughly half the steps, but wrist motion diverged (front_200: right hand much higher), qdist increased from .0026 at step150 to .371 at200. At300 retrieval frame452, qdist .436, vdist .00979. No validated recovery. Episode3 tests gain2 only before retrieved frame300 then gain1 to protect wrist/release timing.
14
+
15
+ Episode 3: gain2 until retrieved frame300 then gain1. Front_200 shows blue ball below right hand during wrist motion, indicating lost grasp; trial ended at250 after reducing interactive limit. Native success false. Faster initial hand deltas are not accepted as an improvement. Plan: restore gain1, independent baseline rehearsal at4; use episodes5-16 to test observation/pose weighting and recovery behavior;17-20 independent exact-version rehearsals.
16
+
17
+ Episode1 audit: actions.npy has 455 proposed actions from native step2 onward, last rejected after terminal. First action reconstructed deterministically from saved initial observation and preserved original offline controller in first_action_reconstructed.npy; marked reconstructed, not contemporaneously logged.
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/freeze_manifest.json ADDED
@@ -0,0 +1,96 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "files": {
3
+ "__pycache__/build_memory.cpython-310.pyc": "a340a6f4053317b248d8bae0fbf444ad7ca0d90e6910f2fe4c8ce221e68f4fcf",
4
+ "__pycache__/check_synthetic.cpython-310.pyc": "7a36b4fb7ecebbd9bd9af75ed37e4351520b8be2d128136d3845d532d04ef2ca",
5
+ "__pycache__/controller.cpython-310.pyc": "35af55a90af49234b1bf22e7bd7c5fae9e5bccaa01a028e151a70aacf4d3ea69",
6
+ "__pycache__/features.cpython-310.pyc": "ea48231dba7b56953fec1128350bab0cd8e2975ce53d81e50c0c59a2af603e8f",
7
+ "build_memory.py": "fd15e74a8d1c05fd1986ef65ac24ad2b5d8199046cd06c72e4b3eb89ea8cbdd8",
8
+ "check_synthetic.py": "080d5753399cc4f91c5bb3763e023b6c260e13e0b11b1164a36cf2f882a534ba",
9
+ "checks.txt": "3e40feb093597670d15454804bdcf7a7ce0a3d0ef3705a7f78de2ea3b29c5058",
10
+ "config.json": "23a9a85db4e641dd7a91b018230436300781579ae3dd5ffb93444dc533701fda",
11
+ "controller.py": "2b2420375df16ce2be06e8ee9543862547b36aa01f220816dd736508bda7c529",
12
+ "demo_manifest.json": "3eb6423dc1164d2f2db7046eea384f668884705e427629f5390cc4f0ec66e650",
13
+ "demonstrations_readonly/demo_00.h5": "e6788fbf5e51d0d75bb3c0b9b63f6e0deea78cc72826ef6b7adb1dfdc9a866cd",
14
+ "demonstrations_readonly/demo_01.h5": "e5c5ccc5750e50aa84d3a162c2d440bc107c700caaf94dc5baf1b8a10da3baff",
15
+ "demonstrations_readonly/demo_02.h5": "4eeb1b7f8ea2f08042656412359036eb1c62f7388240507bc5455947ac156a23",
16
+ "demonstrations_readonly/demo_03.h5": "d56aa11e10fe1dede2e9b29118ff375e9622cb427fc7d41e314b8357979a03ad",
17
+ "demonstrations_readonly/demo_04.h5": "2f3b321cdc749997bf4906e4022684bdd0fe2a4e3773c4bbea740313a410b652",
18
+ "demonstrations_readonly/demo_05.h5": "a654d095a6773246ce04c021941f36800ac984e8d394252c3ff392e446bb2242",
19
+ "demonstrations_readonly/demo_06.h5": "a318cc09e376a66ec514df7b215bc12b5ae15ece64fe4bbe39df3a33252e1dfd",
20
+ "demonstrations_readonly/demo_07.h5": "1e78e3399a4e6d0a6f71547ade517cc428d39c28b5e01a2690550fa3128f51ef",
21
+ "demonstrations_readonly/demo_08.h5": "61e858727b4ea4e2eeff1c4e156c575f39beaeeede80100f5906c98700623d6b",
22
+ "demonstrations_readonly/demo_09.h5": "11feb420b28449d0fb6b4119ca5ecccefe63bc7a6f0a52d7f0bbd7284d4bea4e",
23
+ "demonstrations_readonly/demo_10.h5": "f9914be698a1caebd2b56f34a34a6a808b1f63c9c1328ec54290e2f43fe2d219",
24
+ "demonstrations_readonly/demo_11.h5": "384997aba3ebf7d9942eed1af72cd67a7e3e7c23fe3384beb18f18d7749f2b4a",
25
+ "demonstrations_readonly/demo_12.h5": "15dc0edc5f808474a0169e47048e3863a63439b75d5fc97a14a3d77e6732a967",
26
+ "demonstrations_readonly/demo_13.h5": "44187d6446b3d77edc706f2fed757d24ea644c6fb251824ad033d75b7b012f8b",
27
+ "demonstrations_readonly/demo_14.h5": "b7ebcc1f4bc491c919e87bb001943d19c112d8418147e565f304762f2094d07c",
28
+ "demonstrations_readonly/demo_15.h5": "100b37f92e1eb56c1dbc1e66f5bc6c07a3b82c652aafaa10bcf7b3fb195533b5",
29
+ "demonstrations_readonly/demo_16.h5": "83aedc0863227d57549ec16a9eb1fa5eed3b0af0e1da02e4faf7b80ac575fecb",
30
+ "demonstrations_readonly/demo_17.h5": "b771d3b16928865cf6d1d5ddf172d1521aa039f729544211d9616e34eb244bc0",
31
+ "demonstrations_readonly/demo_18.h5": "6104efa473760e7d71c9ed2e53d3f588f1998c95f40248bc8d0bfc556d15ebfb",
32
+ "demonstrations_readonly/demo_19.h5": "222aae248b23fc50e0180914fa559c65255776af1b030c506245cb730e737931",
33
+ "environment.lock": "f9f11f093744865020243315494d48c69e38ff9e67783e7cf5f0a1e5f22b34aa",
34
+ "features.py": "216c10f6656afb8ac2085e5a9cfe851be2e88d04e66a825449f7f646dd956045",
35
+ "findings.md": "39cdac74a3f47b9adf5a34eeac7ab4a7380b95a57033cc9c3fe5719532353ff6",
36
+ "handoff.md": "068427fed1e7f8545df20f761421259282322e974dd57deb2065038a7dc0413d",
37
+ "memory.npz": "fd6596caf8c231862ef69502a8428cb994249a60d38563ac2757f0f4b280239c",
38
+ "model_tool_manifest.yaml": "c1046861305129f5c810dca6d516f90420f630a2ec0ebfef105e7b77cebdcf2d",
39
+ "policy_runner.py": "cd8a30fc763583fb9bf6570414f01d56aba16241c733fc0cdfdfab0f8f27270e",
40
+ "skills.md": "62038507ac8c52460fd8b738ed8749357b78d1b1e132a0e5a055c0f823f6bb22",
41
+ "worker_api.py": "88cca5fa5c1c9570ca2bd90e8848f67c09224fa498d94e63634ad191e913c41e"
42
+ },
43
+ "bundle_hash": "d309fa881b810013042c6525920f7e397f91b86bdeacd4f2b2287c89d62030f2",
44
+ "exploration_ledger_hash": "b8da681e0b72f85e0eac10fecfd49396a7c25ac4058e6d0505cb15eb4463bd1a",
45
+ "entrypoint": "controller.reset_episode(); controller.act(obs)",
46
+ "selection": {
47
+ "rule": "exact-version success rate, valid rehearsal count, recency",
48
+ "rehearsals": [
49
+ {
50
+ "bundle_hash": "d309fa881b810013042c6525920f7e397f91b86bdeacd4f2b2287c89d62030f2",
51
+ "episode": 4,
52
+ "independent": true,
53
+ "score_source": "native_step_feedback",
54
+ "status": "finished",
55
+ "success": true,
56
+ "valid": true
57
+ },
58
+ {
59
+ "episode": 17,
60
+ "bundle_hash": "dbb4dd9cf2831e3d7eddf967f59a3a0125cf8e4decd1dfea29195a531ab898f3",
61
+ "independent": true,
62
+ "valid": true,
63
+ "status": "exception",
64
+ "success": false,
65
+ "score_source": "native_step_feedback"
66
+ },
67
+ {
68
+ "episode": 18,
69
+ "bundle_hash": "679e517e9ef5bdce53323e2fc43c33f0efdab15130e681e94d1c5927748a13cd",
70
+ "independent": true,
71
+ "valid": true,
72
+ "status": "finished",
73
+ "success": true,
74
+ "score_source": "native_step_feedback"
75
+ },
76
+ {
77
+ "episode": 19,
78
+ "bundle_hash": "679e517e9ef5bdce53323e2fc43c33f0efdab15130e681e94d1c5927748a13cd",
79
+ "independent": true,
80
+ "valid": true,
81
+ "status": "finished",
82
+ "success": false,
83
+ "score_source": "native_step_feedback"
84
+ },
85
+ {
86
+ "episode": 20,
87
+ "bundle_hash": "679e517e9ef5bdce53323e2fc43c33f0efdab15130e681e94d1c5927748a13cd",
88
+ "independent": true,
89
+ "valid": true,
90
+ "status": "finished",
91
+ "success": true,
92
+ "score_source": "native_step_feedback"
93
+ }
94
+ ]
95
+ }
96
+ }
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/handoff.md ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ # E20 candidate handoff
2
+
3
+ Offline runnable controller.py exports reset_episode() and act(obs). Requires bundled memory.npz, features.py, config.json, NumPy and OpenCV. No network, model provider, neural weights, simulator imports, or online development. All memory comes from this task's authorized D20 observations/actions.
4
+
5
+ Runner owns initial hold. Raw observations or observations with step_id supported. Contract 44-component ordering and 1500-step horizon enforced. Missing/invalid observations and extreme retrieval mismatch yield zero deltas. Action gain and observation-retrieved phase threshold are configured in config.json. These modify deltas, not retrieval provenance or reset identity.
6
+
7
+ Episode 1 baseline gain 1 reached native success at step 456. That result is not an independent rehearsal and does not prove three-ball continuous juggling. Visible success requires unambiguous three-ball repeated catch/throw cycles; ambiguity is false. Rehearsal and freeze records outside this bundle are authoritative for exact hash selection. See findings.md for tested limitations.
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/model_tool_manifest.yaml ADDED
@@ -0,0 +1,119 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "stage": "E20 legal exploration",
3
+ "status": "candidate_for_independent_rehearsal",
4
+ "task_id": "DualArmJugglingBall-v1",
5
+ "worker": "Agent-D20-E20-Compiled",
6
+ "neural_finetuning": false,
7
+ "pretrained_weights_loaded": [],
8
+ "frozen_weight_references": {
9
+ "controller": "none; deterministic features and nonparametric example memory",
10
+ "coding_model": "provider-supplied model; exact checkpoint reference unavailable to worker; weights unchanged"
11
+ },
12
+ "tools_used": [
13
+ "functions.exec orchestration",
14
+ "exec_command: local bash and Python",
15
+ "view_image: contact-sheet inspection",
16
+ "worker_api.API.read_demo: authorized local HDF5 access only",
17
+ "Python standard library: importlib, json, pathlib, hashlib, sys, importlib.metadata",
18
+ "NumPy: descriptive statistics, memory arrays, nearest-example distances",
19
+ "OpenCV: RGB resizing, HSV segmentation, connected components, contact sheets",
20
+ "h5py: demonstration reads via API",
21
+ "Trusted worker API: budget_status, next_episode, latest_observation, execute_actions, finish_episode, rehearse, score_rehearsal, freeze",
22
+ "Trusted worker API: budget_status, next_episode, latest_observation, execute_actions, finish_episode, rehearse, score_rehearsal, freeze"
23
+ ],
24
+ "network": {
25
+ "used": false,
26
+ "restriction": "Internet restricted to audited provider proxy; no external queries attempted",
27
+ "proxy_availability": "not tested"
28
+ },
29
+ "versions": {
30
+ "python": "3.10.20 (main, Mar 11 2026, 17:46:40) [GCC 14.3.0]",
31
+ "numpy": "2.2.6",
32
+ "cv2_imported": "4.12.0",
33
+ "h5py": "3.16.0",
34
+ "opencv-python_distribution": "4.13.0.92",
35
+ "opencv-python-headless_distribution": "4.12.0.88"
36
+ },
37
+ "runtime_dependencies": [
38
+ "Python >=3.10",
39
+ "numpy",
40
+ "cv2 (OpenCV)"
41
+ ],
42
+ "offline_dependencies": [
43
+ "h5py",
44
+ "/worker_api.py",
45
+ "authorized /demonstrations",
46
+ "/demo_manifest.json",
47
+ "/task_contract.json"
48
+ ],
49
+ "available_not_used": [
50
+ "SciPy 1.15.3",
51
+ "torch 2.10.0 (not imported; no weights loaded)"
52
+ ],
53
+ "provenance": {
54
+ "inputs": {
55
+ "/task_contract.json": "eca11c3c3f4821f20576055fc0498f22057f1a6a89222f20b283afec73545aeb",
56
+ "/demo_manifest.json": "2af003d315c7a6cd667b814e9d918c1d18711674ea472e386dfaa069585da0f1",
57
+ "/worker_api.py": "88cca5fa5c1c9570ca2bd90e8848f67c09224fa498d94e63634ad191e913c41e"
58
+ },
59
+ "demo_ids": [
60
+ "demo_00",
61
+ "demo_01",
62
+ "demo_02",
63
+ "demo_03",
64
+ "demo_04",
65
+ "demo_05",
66
+ "demo_06",
67
+ "demo_07",
68
+ "demo_08",
69
+ "demo_09",
70
+ "demo_10",
71
+ "demo_11",
72
+ "demo_12",
73
+ "demo_13",
74
+ "demo_14",
75
+ "demo_15",
76
+ "demo_16",
77
+ "demo_17",
78
+ "demo_18",
79
+ "demo_19"
80
+ ],
81
+ "data_access": "Only fields authorized by API.read_demo; all actions/proprio and front/top frames; selected other views for demo_00",
82
+ "visual_inspection_ledger": "../demo_analysis.md",
83
+ "memory": "memory.npz",
84
+ "memory_shapes": {
85
+ "q": [
86
+ 9659,
87
+ 92
88
+ ],
89
+ "v": [
90
+ 9659,
91
+ 1184
92
+ ],
93
+ "a": [
94
+ 9659,
95
+ 44
96
+ ],
97
+ "scale": [
98
+ 92
99
+ ],
100
+ "demo": [
101
+ 9659
102
+ ],
103
+ "frame": [
104
+ 9659
105
+ ]
106
+ },
107
+ "compilation": "build_memory.py; handcrafted feature extraction and descriptive scales only; no gradient training",
108
+ "runner_manifest_and_environment_lock": "supplied separately, not asserted verified here"
109
+ },
110
+ "known_limitations": [
111
+ "Two balls visible in inspected demos versus three-ball continuous juggling instruction",
112
+ "No observable success labels or validated completion detector",
113
+ "No camera extrinsics, root-to-camera transform, robust ball tracking or physical model",
114
+ "Nearest-example selection can stall, switch discontinuously or fail off distribution",
115
+ "Manual color/distance/OOD parameters are uncalibrated",
116
+ "Both OpenCV distributions installed; actual import version recorded",
117
+ "Provider model checkpoint not exposed; no model modification performed"
118
+ ]
119
+ }
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/policy_runner.py ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Frozen-program entry point; no development agent and no reset capability."""
2
+ import importlib.util
3
+ import sys
4
+
5
+ sys.path.insert(0,'/artifact')
6
+ sys.path.insert(0,'/')
7
+ from worker_api import API
8
+
9
+
10
+ def main():
11
+ spec=importlib.util.spec_from_file_location('controller','/artifact/controller.py')
12
+ module=importlib.util.module_from_spec(spec)
13
+ spec.loader.exec_module(module)
14
+ module.reset_episode()
15
+ api=API()
16
+ result=api.latest_observation()
17
+ while not (result['terminated'] or result['truncated'] or result['horizon_reached']):
18
+ obs=dict(result['observation'],step_id=result['step_id'])
19
+ action=module.act(obs)
20
+ if action is None:break
21
+ response=api.execute_actions([action])
22
+ result=response['steps'][-1]
23
+
24
+
25
+ if __name__=='__main__':main()
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/skills.md ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ # Draft capabilities
2
+
3
+ - Read normalized qpos/qvel in the four contract paths and RGB in HWC or singleton-batched HWC format.
4
+ - Extract deterministic color-component and thumbnail features from current front/top cameras.
5
+ - Retrieve a legal joint action from all 20 authorized demonstration memories on each observation.
6
+ - Track consecutive visual differences, reset task-local state, and respect runner-owned hold/horizon accounting.
7
+
8
+ These are implementation capabilities, not validated task skills. No external skill, task-specific expert, simulator, pretrained perception model or neural training was used. No demonstrated three-ball juggling completion criterion is available.
agent_d20_e20/frozen_20261102/tasks/DualArmJugglingBall-v1/frozen/worker_api.py ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Standalone worker client; safe to mount without the runner package."""
2
+ import base64
3
+ import json
4
+ import socket
5
+ import struct
6
+ import uuid
7
+ import os
8
+ from pathlib import Path
9
+
10
+ import numpy as np
11
+
12
+
13
+ def array_hook(value):
14
+ if value.get('__array__') is True:
15
+ dtype=np.dtype(value['dtype'])
16
+ if dtype.kind not in 'bifu':raise ValueError('non-numeric array from service')
17
+ raw=base64.b64decode(value['data'],validate=True)
18
+ return np.frombuffer(raw,dtype=dtype).reshape(value['shape']).copy()
19
+ return value
20
+
21
+
22
+ class API:
23
+ def __init__(self,socket_path='/rpc/control.sock',demo_root='/demonstrations',manifest='/demo_manifest.json'):
24
+ self.socket_path=socket_path
25
+ self.demo_root=Path(demo_root)
26
+ self.manifest=Path(manifest)
27
+
28
+ def call(self,method,arguments=None,request_id=None):
29
+ # Caller can persist a request ID before sending. Never automatically retry.
30
+ request_id=request_id or str(uuid.uuid4())
31
+ request={'request_id':request_id,'method':method,'arguments':arguments or {}}
32
+ if os.environ.get('AGENT_E20_RPC_TOKEN'):
33
+ request['token']=os.environ['AGENT_E20_RPC_TOKEN']
34
+ payload=json.dumps(request,allow_nan=False).encode()
35
+ if len(payload)>65536:raise ValueError('request too large')
36
+ with socket.socket(socket.AF_UNIX,socket.SOCK_STREAM) as s:
37
+ s.settimeout(1810);s.connect(self.socket_path)
38
+ s.sendall(struct.pack('!I',len(payload))+payload)
39
+ def exact(n):
40
+ chunks=[]
41
+ while n:
42
+ chunk=s.recv(min(n,1024*1024))
43
+ if not chunk:raise RuntimeError('RPC interrupted; do not replay action with a new ID')
44
+ chunks.append(chunk);n-=len(chunk)
45
+ return b''.join(chunks)
46
+ length=struct.unpack('!I',exact(4))[0]
47
+ if length>64*1024*1024:raise RuntimeError('response too large')
48
+ result=json.loads(exact(length),object_hook=array_hook)
49
+ if not result['ok']:raise RuntimeError(result['error'])
50
+ return result['result']
51
+
52
+ def next_episode(self,request_id=None):return self.call('next_episode',request_id=request_id)
53
+ def latest_observation(self):return self.call('latest_observation')
54
+ def execute_actions(self,actions,request_id=None):
55
+ return self.call('execute_actions',{'actions':np.asarray(actions).tolist()},request_id)
56
+ def finish_episode(self):return self.call('finish_episode')
57
+ def budget_status(self):return self.call('budget_status')
58
+ def rehearse(self,candidate='candidate',request_id=None):
59
+ return self.call('rehearse',{'candidate':candidate},request_id)
60
+ def score_rehearsal(self,episode,visible_success,evidence):
61
+ return self.call('score_rehearsal',{'episode':episode,'visible_success':visible_success,'evidence':evidence})
62
+ def freeze(self):return self.call('freeze')
63
+
64
+ def read_demo(self,demo_id,start=0,stop=None,fields=None):
65
+ import h5py
66
+ manifest=json.loads(self.manifest.read_text())
67
+ if manifest.get('status') not in ('released','released_for_offline_analysis'):
68
+ raise RuntimeError('demonstrations not released by trusted preflight')
69
+ entries={e['demo_id']:e for e in manifest['episodes']}
70
+ if demo_id not in entries:raise ValueError('unauthorized demo ID')
71
+ entry=entries[demo_id]
72
+ allowed={f['path'] for f in entry['fields']}
73
+ fields=sorted(allowed) if fields is None else fields
74
+ if not set(fields)<=allowed:raise ValueError('unauthorized demo field')
75
+ stop=entry['length']+1 if stop is None else stop
76
+ if not isinstance(start,int) or not isinstance(stop,int) or not 0<=start<stop<=entry['length']+1:
77
+ raise ValueError('invalid frame interval')
78
+ path=(self.demo_root/entry['file']).resolve()
79
+ if path.parent!=self.demo_root.resolve():raise ValueError('invalid demo file')
80
+ with h5py.File(path,'r') as f:
81
+ return {field:f[field][start:min(stop,len(f[field]))] for field in fields}
agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/controller.py ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Legal D20 phase controller with geometric top-camera depth correction."""
2
+ from pathlib import Path
3
+ import numpy as np
4
+ from vision import ball
5
+ from kinematics import fk
6
+ _DB=None
7
+ _demo=None
8
+ _extra=0.
9
+ _last=None
10
+ _prev=None
11
+
12
+ def reset_episode():
13
+ global _demo,_extra,_last,_prev
14
+ _demo=None;_extra=0.;_last=None;_prev=None
15
+
16
+ def act(obs):
17
+ global _DB,_demo,_extra,_last,_prev
18
+ step=int(obs['step_id']);t=step-1
19
+ if step>=210:return None
20
+ if _DB is None:
21
+ with np.load(Path(__file__).with_name('reference.npz')) as f:_DB={k:f[k] for k in f.files if k!='rgb'}
22
+ _DB['ball']=np.load(Path(__file__).with_name('ball_reference.npy'))
23
+ b=ball(obs['rgb']['top_camera'],_prev)
24
+ if b is not None:_prev=b
25
+ if _demo is None:
26
+ starts=np.flatnonzero(_DB['phase']==0)
27
+ j=starts[np.argmin(np.abs(_DB['ball'][starts,0]-(b[0] if b is not None else 240)))]
28
+ _demo=0
29
+ ids=np.flatnonzero(_DB['demo_index']==_demo);j=ids[min(t,len(ids)-1)]
30
+ a=_DB['a'][j].copy();_last=float(min(t,len(ids)-1))
31
+ if t<180 and b is not None:
32
+ target=.00221695*float(b[0])-.530522
33
+ for robot,index,sign in [('panda-0',0,1.),('panda-1',7,-1.)]:
34
+ q=np.asarray(obs['proprioception'][f'agent/{robot}/qpos']).reshape(-1)
35
+ depth=sign*fk(q)[1]
36
+ a[index]=np.clip(5.*(target-depth),-.18,.18)
37
+ if t>=140:
38
+ ax=a[[0,7]].copy();a[:]=0.;a[[6,13]]=-1.
39
+ if t<180:
40
+ a[[0,7]]=ax;a[[2,9]]=.1;a[3]=.03;a[10]=-.03
41
+ for robot,index,sign in [('panda-0',1,-1.),('panda-1',8,1.)]:
42
+ q=np.asarray(obs['proprioception'][f'agent/{robot}/qpos']).reshape(-1)
43
+ a[index]=sign*np.clip(5.*(.58-fk(q)[0]),-.04,.04)
44
+ return np.clip(a,-1,1).astype(np.float32)
agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/demo_manifest.json ADDED
@@ -0,0 +1,2168 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task_id": "DualArmLiftBall-v1",
3
+ "status": "released",
4
+ "selection_rule": "prefer existing pinned D20, otherwise numeric trajectory ID sort before reading data",
5
+ "lineage_status": "pinned_comparator",
6
+ "episodes": [
7
+ {
8
+ "demo_id": "demo_00",
9
+ "file": "demo_00.h5",
10
+ "length": 140,
11
+ "fields": [
12
+ {
13
+ "path": "actions/panda-0",
14
+ "shape": [
15
+ 140,
16
+ 7
17
+ ],
18
+ "dtype": "float32"
19
+ },
20
+ {
21
+ "path": "actions/panda-1",
22
+ "shape": [
23
+ 140,
24
+ 7
25
+ ],
26
+ "dtype": "float32"
27
+ },
28
+ {
29
+ "path": "obs/agent/panda-0/qpos",
30
+ "shape": [
31
+ 141,
32
+ 9
33
+ ],
34
+ "dtype": "float32"
35
+ },
36
+ {
37
+ "path": "obs/agent/panda-0/qvel",
38
+ "shape": [
39
+ 141,
40
+ 9
41
+ ],
42
+ "dtype": "float32"
43
+ },
44
+ {
45
+ "path": "obs/agent/panda-1/qpos",
46
+ "shape": [
47
+ 141,
48
+ 9
49
+ ],
50
+ "dtype": "float32"
51
+ },
52
+ {
53
+ "path": "obs/agent/panda-1/qvel",
54
+ "shape": [
55
+ 141,
56
+ 9
57
+ ],
58
+ "dtype": "float32"
59
+ },
60
+ {
61
+ "path": "obs/sensor_data/front_camera/rgb",
62
+ "shape": [
63
+ 141,
64
+ 240,
65
+ 320,
66
+ 3
67
+ ],
68
+ "dtype": "uint8"
69
+ },
70
+ {
71
+ "path": "obs/sensor_data/left_camera/rgb",
72
+ "shape": [
73
+ 141,
74
+ 240,
75
+ 320,
76
+ 3
77
+ ],
78
+ "dtype": "uint8"
79
+ },
80
+ {
81
+ "path": "obs/sensor_data/robot_camera_0/rgb",
82
+ "shape": [
83
+ 141,
84
+ 240,
85
+ 320,
86
+ 3
87
+ ],
88
+ "dtype": "uint8"
89
+ },
90
+ {
91
+ "path": "obs/sensor_data/robot_camera_1/rgb",
92
+ "shape": [
93
+ 141,
94
+ 240,
95
+ 320,
96
+ 3
97
+ ],
98
+ "dtype": "uint8"
99
+ },
100
+ {
101
+ "path": "obs/sensor_data/top_camera/rgb",
102
+ "shape": [
103
+ 141,
104
+ 320,
105
+ 240,
106
+ 3
107
+ ],
108
+ "dtype": "uint8"
109
+ }
110
+ ],
111
+ "trajectory_sha256": "0ae15cdef87db24bcf5d069394179b3c6275c4c1380104f645cabf87e95d12c3",
112
+ "source_sha256": "a988022aae2e6d19647e5c938412a125545987e733e0aed687f0f327062a90c0",
113
+ "view_sha256": "fc2accfe6d35fa8d7395d00a18969504b4157d7aa67aca75d64baa98b06e57a6"
114
+ },
115
+ {
116
+ "demo_id": "demo_01",
117
+ "file": "demo_01.h5",
118
+ "length": 140,
119
+ "fields": [
120
+ {
121
+ "path": "actions/panda-0",
122
+ "shape": [
123
+ 140,
124
+ 7
125
+ ],
126
+ "dtype": "float32"
127
+ },
128
+ {
129
+ "path": "actions/panda-1",
130
+ "shape": [
131
+ 140,
132
+ 7
133
+ ],
134
+ "dtype": "float32"
135
+ },
136
+ {
137
+ "path": "obs/agent/panda-0/qpos",
138
+ "shape": [
139
+ 141,
140
+ 9
141
+ ],
142
+ "dtype": "float32"
143
+ },
144
+ {
145
+ "path": "obs/agent/panda-0/qvel",
146
+ "shape": [
147
+ 141,
148
+ 9
149
+ ],
150
+ "dtype": "float32"
151
+ },
152
+ {
153
+ "path": "obs/agent/panda-1/qpos",
154
+ "shape": [
155
+ 141,
156
+ 9
157
+ ],
158
+ "dtype": "float32"
159
+ },
160
+ {
161
+ "path": "obs/agent/panda-1/qvel",
162
+ "shape": [
163
+ 141,
164
+ 9
165
+ ],
166
+ "dtype": "float32"
167
+ },
168
+ {
169
+ "path": "obs/sensor_data/front_camera/rgb",
170
+ "shape": [
171
+ 141,
172
+ 240,
173
+ 320,
174
+ 3
175
+ ],
176
+ "dtype": "uint8"
177
+ },
178
+ {
179
+ "path": "obs/sensor_data/left_camera/rgb",
180
+ "shape": [
181
+ 141,
182
+ 240,
183
+ 320,
184
+ 3
185
+ ],
186
+ "dtype": "uint8"
187
+ },
188
+ {
189
+ "path": "obs/sensor_data/robot_camera_0/rgb",
190
+ "shape": [
191
+ 141,
192
+ 240,
193
+ 320,
194
+ 3
195
+ ],
196
+ "dtype": "uint8"
197
+ },
198
+ {
199
+ "path": "obs/sensor_data/robot_camera_1/rgb",
200
+ "shape": [
201
+ 141,
202
+ 240,
203
+ 320,
204
+ 3
205
+ ],
206
+ "dtype": "uint8"
207
+ },
208
+ {
209
+ "path": "obs/sensor_data/top_camera/rgb",
210
+ "shape": [
211
+ 141,
212
+ 320,
213
+ 240,
214
+ 3
215
+ ],
216
+ "dtype": "uint8"
217
+ }
218
+ ],
219
+ "trajectory_sha256": "2c02e09c7ee630b6adf295e5214bd635981add00e5db4866618cc7e8d42cd5a7",
220
+ "source_sha256": "18b83de5405e4528a857c263e067d0adbb0630231313f93df6b1914b3882a3d9",
221
+ "view_sha256": "564a89cf86c685206ab6442adaee4b7d3a67e59805c97bc37d0d8a2a9e43f8ec"
222
+ },
223
+ {
224
+ "demo_id": "demo_02",
225
+ "file": "demo_02.h5",
226
+ "length": 138,
227
+ "fields": [
228
+ {
229
+ "path": "actions/panda-0",
230
+ "shape": [
231
+ 138,
232
+ 7
233
+ ],
234
+ "dtype": "float32"
235
+ },
236
+ {
237
+ "path": "actions/panda-1",
238
+ "shape": [
239
+ 138,
240
+ 7
241
+ ],
242
+ "dtype": "float32"
243
+ },
244
+ {
245
+ "path": "obs/agent/panda-0/qpos",
246
+ "shape": [
247
+ 139,
248
+ 9
249
+ ],
250
+ "dtype": "float32"
251
+ },
252
+ {
253
+ "path": "obs/agent/panda-0/qvel",
254
+ "shape": [
255
+ 139,
256
+ 9
257
+ ],
258
+ "dtype": "float32"
259
+ },
260
+ {
261
+ "path": "obs/agent/panda-1/qpos",
262
+ "shape": [
263
+ 139,
264
+ 9
265
+ ],
266
+ "dtype": "float32"
267
+ },
268
+ {
269
+ "path": "obs/agent/panda-1/qvel",
270
+ "shape": [
271
+ 139,
272
+ 9
273
+ ],
274
+ "dtype": "float32"
275
+ },
276
+ {
277
+ "path": "obs/sensor_data/front_camera/rgb",
278
+ "shape": [
279
+ 139,
280
+ 240,
281
+ 320,
282
+ 3
283
+ ],
284
+ "dtype": "uint8"
285
+ },
286
+ {
287
+ "path": "obs/sensor_data/left_camera/rgb",
288
+ "shape": [
289
+ 139,
290
+ 240,
291
+ 320,
292
+ 3
293
+ ],
294
+ "dtype": "uint8"
295
+ },
296
+ {
297
+ "path": "obs/sensor_data/robot_camera_0/rgb",
298
+ "shape": [
299
+ 139,
300
+ 240,
301
+ 320,
302
+ 3
303
+ ],
304
+ "dtype": "uint8"
305
+ },
306
+ {
307
+ "path": "obs/sensor_data/robot_camera_1/rgb",
308
+ "shape": [
309
+ 139,
310
+ 240,
311
+ 320,
312
+ 3
313
+ ],
314
+ "dtype": "uint8"
315
+ },
316
+ {
317
+ "path": "obs/sensor_data/top_camera/rgb",
318
+ "shape": [
319
+ 139,
320
+ 320,
321
+ 240,
322
+ 3
323
+ ],
324
+ "dtype": "uint8"
325
+ }
326
+ ],
327
+ "trajectory_sha256": "206b30b8c36dd2e971edada8ad2fdac6c41660cde39d53609252d96056164afa",
328
+ "source_sha256": "b08ddfe184b82c371ff75e675333f2f398ecd5b2955cb0794d9564b14309dc12",
329
+ "view_sha256": "d3109687e72305fa5f6b97bf4805dc2260072d0197018c9299d9344282e872c1"
330
+ },
331
+ {
332
+ "demo_id": "demo_03",
333
+ "file": "demo_03.h5",
334
+ "length": 140,
335
+ "fields": [
336
+ {
337
+ "path": "actions/panda-0",
338
+ "shape": [
339
+ 140,
340
+ 7
341
+ ],
342
+ "dtype": "float32"
343
+ },
344
+ {
345
+ "path": "actions/panda-1",
346
+ "shape": [
347
+ 140,
348
+ 7
349
+ ],
350
+ "dtype": "float32"
351
+ },
352
+ {
353
+ "path": "obs/agent/panda-0/qpos",
354
+ "shape": [
355
+ 141,
356
+ 9
357
+ ],
358
+ "dtype": "float32"
359
+ },
360
+ {
361
+ "path": "obs/agent/panda-0/qvel",
362
+ "shape": [
363
+ 141,
364
+ 9
365
+ ],
366
+ "dtype": "float32"
367
+ },
368
+ {
369
+ "path": "obs/agent/panda-1/qpos",
370
+ "shape": [
371
+ 141,
372
+ 9
373
+ ],
374
+ "dtype": "float32"
375
+ },
376
+ {
377
+ "path": "obs/agent/panda-1/qvel",
378
+ "shape": [
379
+ 141,
380
+ 9
381
+ ],
382
+ "dtype": "float32"
383
+ },
384
+ {
385
+ "path": "obs/sensor_data/front_camera/rgb",
386
+ "shape": [
387
+ 141,
388
+ 240,
389
+ 320,
390
+ 3
391
+ ],
392
+ "dtype": "uint8"
393
+ },
394
+ {
395
+ "path": "obs/sensor_data/left_camera/rgb",
396
+ "shape": [
397
+ 141,
398
+ 240,
399
+ 320,
400
+ 3
401
+ ],
402
+ "dtype": "uint8"
403
+ },
404
+ {
405
+ "path": "obs/sensor_data/robot_camera_0/rgb",
406
+ "shape": [
407
+ 141,
408
+ 240,
409
+ 320,
410
+ 3
411
+ ],
412
+ "dtype": "uint8"
413
+ },
414
+ {
415
+ "path": "obs/sensor_data/robot_camera_1/rgb",
416
+ "shape": [
417
+ 141,
418
+ 240,
419
+ 320,
420
+ 3
421
+ ],
422
+ "dtype": "uint8"
423
+ },
424
+ {
425
+ "path": "obs/sensor_data/top_camera/rgb",
426
+ "shape": [
427
+ 141,
428
+ 320,
429
+ 240,
430
+ 3
431
+ ],
432
+ "dtype": "uint8"
433
+ }
434
+ ],
435
+ "trajectory_sha256": "8656453da3e3deb57abbe3eb7fc8cc59862ff27c05190d35d3460dc85b89825d",
436
+ "source_sha256": "84cb26fda8bf384d4ff4936ff9806d27ba5ec446f0f493e260ebe2979cc9ed72",
437
+ "view_sha256": "44e8f69c7f533044cc782e9b69b74f4f2fe0c6bca2ef2cd2a987f263125e4a22"
438
+ },
439
+ {
440
+ "demo_id": "demo_04",
441
+ "file": "demo_04.h5",
442
+ "length": 140,
443
+ "fields": [
444
+ {
445
+ "path": "actions/panda-0",
446
+ "shape": [
447
+ 140,
448
+ 7
449
+ ],
450
+ "dtype": "float32"
451
+ },
452
+ {
453
+ "path": "actions/panda-1",
454
+ "shape": [
455
+ 140,
456
+ 7
457
+ ],
458
+ "dtype": "float32"
459
+ },
460
+ {
461
+ "path": "obs/agent/panda-0/qpos",
462
+ "shape": [
463
+ 141,
464
+ 9
465
+ ],
466
+ "dtype": "float32"
467
+ },
468
+ {
469
+ "path": "obs/agent/panda-0/qvel",
470
+ "shape": [
471
+ 141,
472
+ 9
473
+ ],
474
+ "dtype": "float32"
475
+ },
476
+ {
477
+ "path": "obs/agent/panda-1/qpos",
478
+ "shape": [
479
+ 141,
480
+ 9
481
+ ],
482
+ "dtype": "float32"
483
+ },
484
+ {
485
+ "path": "obs/agent/panda-1/qvel",
486
+ "shape": [
487
+ 141,
488
+ 9
489
+ ],
490
+ "dtype": "float32"
491
+ },
492
+ {
493
+ "path": "obs/sensor_data/front_camera/rgb",
494
+ "shape": [
495
+ 141,
496
+ 240,
497
+ 320,
498
+ 3
499
+ ],
500
+ "dtype": "uint8"
501
+ },
502
+ {
503
+ "path": "obs/sensor_data/left_camera/rgb",
504
+ "shape": [
505
+ 141,
506
+ 240,
507
+ 320,
508
+ 3
509
+ ],
510
+ "dtype": "uint8"
511
+ },
512
+ {
513
+ "path": "obs/sensor_data/robot_camera_0/rgb",
514
+ "shape": [
515
+ 141,
516
+ 240,
517
+ 320,
518
+ 3
519
+ ],
520
+ "dtype": "uint8"
521
+ },
522
+ {
523
+ "path": "obs/sensor_data/robot_camera_1/rgb",
524
+ "shape": [
525
+ 141,
526
+ 240,
527
+ 320,
528
+ 3
529
+ ],
530
+ "dtype": "uint8"
531
+ },
532
+ {
533
+ "path": "obs/sensor_data/top_camera/rgb",
534
+ "shape": [
535
+ 141,
536
+ 320,
537
+ 240,
538
+ 3
539
+ ],
540
+ "dtype": "uint8"
541
+ }
542
+ ],
543
+ "trajectory_sha256": "bf5b13743f0575fcb7463c0545f7e4c0833c7d3b28aa63ba47d2f7d89c962b8e",
544
+ "source_sha256": "6af0e7bec4a0014a5d8b9c7a1795304732ab28c7d5e27f68f5119b4fb60e0cf3",
545
+ "view_sha256": "5f7505253cbdbfcb343f05d4a1a0ce694ba54e28a43e6b361af32a9f29f99d17"
546
+ },
547
+ {
548
+ "demo_id": "demo_05",
549
+ "file": "demo_05.h5",
550
+ "length": 140,
551
+ "fields": [
552
+ {
553
+ "path": "actions/panda-0",
554
+ "shape": [
555
+ 140,
556
+ 7
557
+ ],
558
+ "dtype": "float32"
559
+ },
560
+ {
561
+ "path": "actions/panda-1",
562
+ "shape": [
563
+ 140,
564
+ 7
565
+ ],
566
+ "dtype": "float32"
567
+ },
568
+ {
569
+ "path": "obs/agent/panda-0/qpos",
570
+ "shape": [
571
+ 141,
572
+ 9
573
+ ],
574
+ "dtype": "float32"
575
+ },
576
+ {
577
+ "path": "obs/agent/panda-0/qvel",
578
+ "shape": [
579
+ 141,
580
+ 9
581
+ ],
582
+ "dtype": "float32"
583
+ },
584
+ {
585
+ "path": "obs/agent/panda-1/qpos",
586
+ "shape": [
587
+ 141,
588
+ 9
589
+ ],
590
+ "dtype": "float32"
591
+ },
592
+ {
593
+ "path": "obs/agent/panda-1/qvel",
594
+ "shape": [
595
+ 141,
596
+ 9
597
+ ],
598
+ "dtype": "float32"
599
+ },
600
+ {
601
+ "path": "obs/sensor_data/front_camera/rgb",
602
+ "shape": [
603
+ 141,
604
+ 240,
605
+ 320,
606
+ 3
607
+ ],
608
+ "dtype": "uint8"
609
+ },
610
+ {
611
+ "path": "obs/sensor_data/left_camera/rgb",
612
+ "shape": [
613
+ 141,
614
+ 240,
615
+ 320,
616
+ 3
617
+ ],
618
+ "dtype": "uint8"
619
+ },
620
+ {
621
+ "path": "obs/sensor_data/robot_camera_0/rgb",
622
+ "shape": [
623
+ 141,
624
+ 240,
625
+ 320,
626
+ 3
627
+ ],
628
+ "dtype": "uint8"
629
+ },
630
+ {
631
+ "path": "obs/sensor_data/robot_camera_1/rgb",
632
+ "shape": [
633
+ 141,
634
+ 240,
635
+ 320,
636
+ 3
637
+ ],
638
+ "dtype": "uint8"
639
+ },
640
+ {
641
+ "path": "obs/sensor_data/top_camera/rgb",
642
+ "shape": [
643
+ 141,
644
+ 320,
645
+ 240,
646
+ 3
647
+ ],
648
+ "dtype": "uint8"
649
+ }
650
+ ],
651
+ "trajectory_sha256": "26fd2af075dfa50364c3ac4bde4f160f39299dcff8855ace2e38674f28903094",
652
+ "source_sha256": "bb7fca500747c7229797c604e292c34b08ce434239d2caefcefbfa87d6c2ccbc",
653
+ "view_sha256": "e90836977e0cdae992916d6f25e8be7e9093bf20f14df45fc10ad9bf66437b24"
654
+ },
655
+ {
656
+ "demo_id": "demo_06",
657
+ "file": "demo_06.h5",
658
+ "length": 140,
659
+ "fields": [
660
+ {
661
+ "path": "actions/panda-0",
662
+ "shape": [
663
+ 140,
664
+ 7
665
+ ],
666
+ "dtype": "float32"
667
+ },
668
+ {
669
+ "path": "actions/panda-1",
670
+ "shape": [
671
+ 140,
672
+ 7
673
+ ],
674
+ "dtype": "float32"
675
+ },
676
+ {
677
+ "path": "obs/agent/panda-0/qpos",
678
+ "shape": [
679
+ 141,
680
+ 9
681
+ ],
682
+ "dtype": "float32"
683
+ },
684
+ {
685
+ "path": "obs/agent/panda-0/qvel",
686
+ "shape": [
687
+ 141,
688
+ 9
689
+ ],
690
+ "dtype": "float32"
691
+ },
692
+ {
693
+ "path": "obs/agent/panda-1/qpos",
694
+ "shape": [
695
+ 141,
696
+ 9
697
+ ],
698
+ "dtype": "float32"
699
+ },
700
+ {
701
+ "path": "obs/agent/panda-1/qvel",
702
+ "shape": [
703
+ 141,
704
+ 9
705
+ ],
706
+ "dtype": "float32"
707
+ },
708
+ {
709
+ "path": "obs/sensor_data/front_camera/rgb",
710
+ "shape": [
711
+ 141,
712
+ 240,
713
+ 320,
714
+ 3
715
+ ],
716
+ "dtype": "uint8"
717
+ },
718
+ {
719
+ "path": "obs/sensor_data/left_camera/rgb",
720
+ "shape": [
721
+ 141,
722
+ 240,
723
+ 320,
724
+ 3
725
+ ],
726
+ "dtype": "uint8"
727
+ },
728
+ {
729
+ "path": "obs/sensor_data/robot_camera_0/rgb",
730
+ "shape": [
731
+ 141,
732
+ 240,
733
+ 320,
734
+ 3
735
+ ],
736
+ "dtype": "uint8"
737
+ },
738
+ {
739
+ "path": "obs/sensor_data/robot_camera_1/rgb",
740
+ "shape": [
741
+ 141,
742
+ 240,
743
+ 320,
744
+ 3
745
+ ],
746
+ "dtype": "uint8"
747
+ },
748
+ {
749
+ "path": "obs/sensor_data/top_camera/rgb",
750
+ "shape": [
751
+ 141,
752
+ 320,
753
+ 240,
754
+ 3
755
+ ],
756
+ "dtype": "uint8"
757
+ }
758
+ ],
759
+ "trajectory_sha256": "41c1115be80e39d10232092f5419891fec8888d1846a59bd3a91e3dc1948bb25",
760
+ "source_sha256": "ccb59adbf3b7a93cc5213f99e5c5966903e0262995ac94f46ab08a976c57e0e2",
761
+ "view_sha256": "da208e25be7a82564362249ad8247b88acb9321dba9b179ff0a99c830b14b96d"
762
+ },
763
+ {
764
+ "demo_id": "demo_07",
765
+ "file": "demo_07.h5",
766
+ "length": 140,
767
+ "fields": [
768
+ {
769
+ "path": "actions/panda-0",
770
+ "shape": [
771
+ 140,
772
+ 7
773
+ ],
774
+ "dtype": "float32"
775
+ },
776
+ {
777
+ "path": "actions/panda-1",
778
+ "shape": [
779
+ 140,
780
+ 7
781
+ ],
782
+ "dtype": "float32"
783
+ },
784
+ {
785
+ "path": "obs/agent/panda-0/qpos",
786
+ "shape": [
787
+ 141,
788
+ 9
789
+ ],
790
+ "dtype": "float32"
791
+ },
792
+ {
793
+ "path": "obs/agent/panda-0/qvel",
794
+ "shape": [
795
+ 141,
796
+ 9
797
+ ],
798
+ "dtype": "float32"
799
+ },
800
+ {
801
+ "path": "obs/agent/panda-1/qpos",
802
+ "shape": [
803
+ 141,
804
+ 9
805
+ ],
806
+ "dtype": "float32"
807
+ },
808
+ {
809
+ "path": "obs/agent/panda-1/qvel",
810
+ "shape": [
811
+ 141,
812
+ 9
813
+ ],
814
+ "dtype": "float32"
815
+ },
816
+ {
817
+ "path": "obs/sensor_data/front_camera/rgb",
818
+ "shape": [
819
+ 141,
820
+ 240,
821
+ 320,
822
+ 3
823
+ ],
824
+ "dtype": "uint8"
825
+ },
826
+ {
827
+ "path": "obs/sensor_data/left_camera/rgb",
828
+ "shape": [
829
+ 141,
830
+ 240,
831
+ 320,
832
+ 3
833
+ ],
834
+ "dtype": "uint8"
835
+ },
836
+ {
837
+ "path": "obs/sensor_data/robot_camera_0/rgb",
838
+ "shape": [
839
+ 141,
840
+ 240,
841
+ 320,
842
+ 3
843
+ ],
844
+ "dtype": "uint8"
845
+ },
846
+ {
847
+ "path": "obs/sensor_data/robot_camera_1/rgb",
848
+ "shape": [
849
+ 141,
850
+ 240,
851
+ 320,
852
+ 3
853
+ ],
854
+ "dtype": "uint8"
855
+ },
856
+ {
857
+ "path": "obs/sensor_data/top_camera/rgb",
858
+ "shape": [
859
+ 141,
860
+ 320,
861
+ 240,
862
+ 3
863
+ ],
864
+ "dtype": "uint8"
865
+ }
866
+ ],
867
+ "trajectory_sha256": "f50458b8be6a85fa663551122ab9dd51d0805f5a35dbe657e6dd69f4cd873456",
868
+ "source_sha256": "8fd230b1f25f04cd7b6c0998d6ec7e6e3e267bbca466d81a4f112d55aaeac2c0",
869
+ "view_sha256": "88b6bcf7a83f9fbf473820b41f71f1df19279efe9f0ed46a5f78ba5c93928f9e"
870
+ },
871
+ {
872
+ "demo_id": "demo_08",
873
+ "file": "demo_08.h5",
874
+ "length": 145,
875
+ "fields": [
876
+ {
877
+ "path": "actions/panda-0",
878
+ "shape": [
879
+ 145,
880
+ 7
881
+ ],
882
+ "dtype": "float32"
883
+ },
884
+ {
885
+ "path": "actions/panda-1",
886
+ "shape": [
887
+ 145,
888
+ 7
889
+ ],
890
+ "dtype": "float32"
891
+ },
892
+ {
893
+ "path": "obs/agent/panda-0/qpos",
894
+ "shape": [
895
+ 146,
896
+ 9
897
+ ],
898
+ "dtype": "float32"
899
+ },
900
+ {
901
+ "path": "obs/agent/panda-0/qvel",
902
+ "shape": [
903
+ 146,
904
+ 9
905
+ ],
906
+ "dtype": "float32"
907
+ },
908
+ {
909
+ "path": "obs/agent/panda-1/qpos",
910
+ "shape": [
911
+ 146,
912
+ 9
913
+ ],
914
+ "dtype": "float32"
915
+ },
916
+ {
917
+ "path": "obs/agent/panda-1/qvel",
918
+ "shape": [
919
+ 146,
920
+ 9
921
+ ],
922
+ "dtype": "float32"
923
+ },
924
+ {
925
+ "path": "obs/sensor_data/front_camera/rgb",
926
+ "shape": [
927
+ 146,
928
+ 240,
929
+ 320,
930
+ 3
931
+ ],
932
+ "dtype": "uint8"
933
+ },
934
+ {
935
+ "path": "obs/sensor_data/left_camera/rgb",
936
+ "shape": [
937
+ 146,
938
+ 240,
939
+ 320,
940
+ 3
941
+ ],
942
+ "dtype": "uint8"
943
+ },
944
+ {
945
+ "path": "obs/sensor_data/robot_camera_0/rgb",
946
+ "shape": [
947
+ 146,
948
+ 240,
949
+ 320,
950
+ 3
951
+ ],
952
+ "dtype": "uint8"
953
+ },
954
+ {
955
+ "path": "obs/sensor_data/robot_camera_1/rgb",
956
+ "shape": [
957
+ 146,
958
+ 240,
959
+ 320,
960
+ 3
961
+ ],
962
+ "dtype": "uint8"
963
+ },
964
+ {
965
+ "path": "obs/sensor_data/top_camera/rgb",
966
+ "shape": [
967
+ 146,
968
+ 320,
969
+ 240,
970
+ 3
971
+ ],
972
+ "dtype": "uint8"
973
+ }
974
+ ],
975
+ "trajectory_sha256": "bfd615b9822103cd385e381495e57f7e455aa2b49639b08376e3feb27ee73989",
976
+ "source_sha256": "dfb6edf712ce184710096523769145dd7569f63dd9205d7eb4dc32884caa4aeb",
977
+ "view_sha256": "f25ee42499e3548463aab47927281d5199527dc19dd4ea5f134f6f583164619b"
978
+ },
979
+ {
980
+ "demo_id": "demo_09",
981
+ "file": "demo_09.h5",
982
+ "length": 140,
983
+ "fields": [
984
+ {
985
+ "path": "actions/panda-0",
986
+ "shape": [
987
+ 140,
988
+ 7
989
+ ],
990
+ "dtype": "float32"
991
+ },
992
+ {
993
+ "path": "actions/panda-1",
994
+ "shape": [
995
+ 140,
996
+ 7
997
+ ],
998
+ "dtype": "float32"
999
+ },
1000
+ {
1001
+ "path": "obs/agent/panda-0/qpos",
1002
+ "shape": [
1003
+ 141,
1004
+ 9
1005
+ ],
1006
+ "dtype": "float32"
1007
+ },
1008
+ {
1009
+ "path": "obs/agent/panda-0/qvel",
1010
+ "shape": [
1011
+ 141,
1012
+ 9
1013
+ ],
1014
+ "dtype": "float32"
1015
+ },
1016
+ {
1017
+ "path": "obs/agent/panda-1/qpos",
1018
+ "shape": [
1019
+ 141,
1020
+ 9
1021
+ ],
1022
+ "dtype": "float32"
1023
+ },
1024
+ {
1025
+ "path": "obs/agent/panda-1/qvel",
1026
+ "shape": [
1027
+ 141,
1028
+ 9
1029
+ ],
1030
+ "dtype": "float32"
1031
+ },
1032
+ {
1033
+ "path": "obs/sensor_data/front_camera/rgb",
1034
+ "shape": [
1035
+ 141,
1036
+ 240,
1037
+ 320,
1038
+ 3
1039
+ ],
1040
+ "dtype": "uint8"
1041
+ },
1042
+ {
1043
+ "path": "obs/sensor_data/left_camera/rgb",
1044
+ "shape": [
1045
+ 141,
1046
+ 240,
1047
+ 320,
1048
+ 3
1049
+ ],
1050
+ "dtype": "uint8"
1051
+ },
1052
+ {
1053
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1054
+ "shape": [
1055
+ 141,
1056
+ 240,
1057
+ 320,
1058
+ 3
1059
+ ],
1060
+ "dtype": "uint8"
1061
+ },
1062
+ {
1063
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1064
+ "shape": [
1065
+ 141,
1066
+ 240,
1067
+ 320,
1068
+ 3
1069
+ ],
1070
+ "dtype": "uint8"
1071
+ },
1072
+ {
1073
+ "path": "obs/sensor_data/top_camera/rgb",
1074
+ "shape": [
1075
+ 141,
1076
+ 320,
1077
+ 240,
1078
+ 3
1079
+ ],
1080
+ "dtype": "uint8"
1081
+ }
1082
+ ],
1083
+ "trajectory_sha256": "6507a7d7eea71affe008d300a7b871b83e0aeef7b1774aa7cef11ed31c3b713c",
1084
+ "source_sha256": "f91837505413989229e45afb86ea34d4028c894cdf2f6f1cad9edc0455567d95",
1085
+ "view_sha256": "fa3eb0447d6b9741bce6057773220476551678883346bc5cd2a438b6b0deac40"
1086
+ },
1087
+ {
1088
+ "demo_id": "demo_10",
1089
+ "file": "demo_10.h5",
1090
+ "length": 139,
1091
+ "fields": [
1092
+ {
1093
+ "path": "actions/panda-0",
1094
+ "shape": [
1095
+ 139,
1096
+ 7
1097
+ ],
1098
+ "dtype": "float32"
1099
+ },
1100
+ {
1101
+ "path": "actions/panda-1",
1102
+ "shape": [
1103
+ 139,
1104
+ 7
1105
+ ],
1106
+ "dtype": "float32"
1107
+ },
1108
+ {
1109
+ "path": "obs/agent/panda-0/qpos",
1110
+ "shape": [
1111
+ 140,
1112
+ 9
1113
+ ],
1114
+ "dtype": "float32"
1115
+ },
1116
+ {
1117
+ "path": "obs/agent/panda-0/qvel",
1118
+ "shape": [
1119
+ 140,
1120
+ 9
1121
+ ],
1122
+ "dtype": "float32"
1123
+ },
1124
+ {
1125
+ "path": "obs/agent/panda-1/qpos",
1126
+ "shape": [
1127
+ 140,
1128
+ 9
1129
+ ],
1130
+ "dtype": "float32"
1131
+ },
1132
+ {
1133
+ "path": "obs/agent/panda-1/qvel",
1134
+ "shape": [
1135
+ 140,
1136
+ 9
1137
+ ],
1138
+ "dtype": "float32"
1139
+ },
1140
+ {
1141
+ "path": "obs/sensor_data/front_camera/rgb",
1142
+ "shape": [
1143
+ 140,
1144
+ 240,
1145
+ 320,
1146
+ 3
1147
+ ],
1148
+ "dtype": "uint8"
1149
+ },
1150
+ {
1151
+ "path": "obs/sensor_data/left_camera/rgb",
1152
+ "shape": [
1153
+ 140,
1154
+ 240,
1155
+ 320,
1156
+ 3
1157
+ ],
1158
+ "dtype": "uint8"
1159
+ },
1160
+ {
1161
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1162
+ "shape": [
1163
+ 140,
1164
+ 240,
1165
+ 320,
1166
+ 3
1167
+ ],
1168
+ "dtype": "uint8"
1169
+ },
1170
+ {
1171
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1172
+ "shape": [
1173
+ 140,
1174
+ 240,
1175
+ 320,
1176
+ 3
1177
+ ],
1178
+ "dtype": "uint8"
1179
+ },
1180
+ {
1181
+ "path": "obs/sensor_data/top_camera/rgb",
1182
+ "shape": [
1183
+ 140,
1184
+ 320,
1185
+ 240,
1186
+ 3
1187
+ ],
1188
+ "dtype": "uint8"
1189
+ }
1190
+ ],
1191
+ "trajectory_sha256": "278a24a25b980af009d6b528af55e2380e05f7540a9ce81c84096caad137da04",
1192
+ "source_sha256": "5d48c538d1b69c26a235064a436b0ce06bfa56670e51daa47c5a6bbb808e469f",
1193
+ "view_sha256": "7eb12dd94e5c0b88e933fa049293bd77929845f044d6d73766f6a3ffcbe950ae"
1194
+ },
1195
+ {
1196
+ "demo_id": "demo_11",
1197
+ "file": "demo_11.h5",
1198
+ "length": 138,
1199
+ "fields": [
1200
+ {
1201
+ "path": "actions/panda-0",
1202
+ "shape": [
1203
+ 138,
1204
+ 7
1205
+ ],
1206
+ "dtype": "float32"
1207
+ },
1208
+ {
1209
+ "path": "actions/panda-1",
1210
+ "shape": [
1211
+ 138,
1212
+ 7
1213
+ ],
1214
+ "dtype": "float32"
1215
+ },
1216
+ {
1217
+ "path": "obs/agent/panda-0/qpos",
1218
+ "shape": [
1219
+ 139,
1220
+ 9
1221
+ ],
1222
+ "dtype": "float32"
1223
+ },
1224
+ {
1225
+ "path": "obs/agent/panda-0/qvel",
1226
+ "shape": [
1227
+ 139,
1228
+ 9
1229
+ ],
1230
+ "dtype": "float32"
1231
+ },
1232
+ {
1233
+ "path": "obs/agent/panda-1/qpos",
1234
+ "shape": [
1235
+ 139,
1236
+ 9
1237
+ ],
1238
+ "dtype": "float32"
1239
+ },
1240
+ {
1241
+ "path": "obs/agent/panda-1/qvel",
1242
+ "shape": [
1243
+ 139,
1244
+ 9
1245
+ ],
1246
+ "dtype": "float32"
1247
+ },
1248
+ {
1249
+ "path": "obs/sensor_data/front_camera/rgb",
1250
+ "shape": [
1251
+ 139,
1252
+ 240,
1253
+ 320,
1254
+ 3
1255
+ ],
1256
+ "dtype": "uint8"
1257
+ },
1258
+ {
1259
+ "path": "obs/sensor_data/left_camera/rgb",
1260
+ "shape": [
1261
+ 139,
1262
+ 240,
1263
+ 320,
1264
+ 3
1265
+ ],
1266
+ "dtype": "uint8"
1267
+ },
1268
+ {
1269
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1270
+ "shape": [
1271
+ 139,
1272
+ 240,
1273
+ 320,
1274
+ 3
1275
+ ],
1276
+ "dtype": "uint8"
1277
+ },
1278
+ {
1279
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1280
+ "shape": [
1281
+ 139,
1282
+ 240,
1283
+ 320,
1284
+ 3
1285
+ ],
1286
+ "dtype": "uint8"
1287
+ },
1288
+ {
1289
+ "path": "obs/sensor_data/top_camera/rgb",
1290
+ "shape": [
1291
+ 139,
1292
+ 320,
1293
+ 240,
1294
+ 3
1295
+ ],
1296
+ "dtype": "uint8"
1297
+ }
1298
+ ],
1299
+ "trajectory_sha256": "365ab66b52d4cc1d4797177dc8f18ed5add6ed5ffbd07aaff4053dddf200eedc",
1300
+ "source_sha256": "682d6601dbd0ad786a16707d40e8eef90cefa0554bd6042cc3b59316aaa6c9ac",
1301
+ "view_sha256": "b8a82cb74ab3edc42993c206c3994936fb53697a9ab685659bf5c790b83a9913"
1302
+ },
1303
+ {
1304
+ "demo_id": "demo_12",
1305
+ "file": "demo_12.h5",
1306
+ "length": 140,
1307
+ "fields": [
1308
+ {
1309
+ "path": "actions/panda-0",
1310
+ "shape": [
1311
+ 140,
1312
+ 7
1313
+ ],
1314
+ "dtype": "float32"
1315
+ },
1316
+ {
1317
+ "path": "actions/panda-1",
1318
+ "shape": [
1319
+ 140,
1320
+ 7
1321
+ ],
1322
+ "dtype": "float32"
1323
+ },
1324
+ {
1325
+ "path": "obs/agent/panda-0/qpos",
1326
+ "shape": [
1327
+ 141,
1328
+ 9
1329
+ ],
1330
+ "dtype": "float32"
1331
+ },
1332
+ {
1333
+ "path": "obs/agent/panda-0/qvel",
1334
+ "shape": [
1335
+ 141,
1336
+ 9
1337
+ ],
1338
+ "dtype": "float32"
1339
+ },
1340
+ {
1341
+ "path": "obs/agent/panda-1/qpos",
1342
+ "shape": [
1343
+ 141,
1344
+ 9
1345
+ ],
1346
+ "dtype": "float32"
1347
+ },
1348
+ {
1349
+ "path": "obs/agent/panda-1/qvel",
1350
+ "shape": [
1351
+ 141,
1352
+ 9
1353
+ ],
1354
+ "dtype": "float32"
1355
+ },
1356
+ {
1357
+ "path": "obs/sensor_data/front_camera/rgb",
1358
+ "shape": [
1359
+ 141,
1360
+ 240,
1361
+ 320,
1362
+ 3
1363
+ ],
1364
+ "dtype": "uint8"
1365
+ },
1366
+ {
1367
+ "path": "obs/sensor_data/left_camera/rgb",
1368
+ "shape": [
1369
+ 141,
1370
+ 240,
1371
+ 320,
1372
+ 3
1373
+ ],
1374
+ "dtype": "uint8"
1375
+ },
1376
+ {
1377
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1378
+ "shape": [
1379
+ 141,
1380
+ 240,
1381
+ 320,
1382
+ 3
1383
+ ],
1384
+ "dtype": "uint8"
1385
+ },
1386
+ {
1387
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1388
+ "shape": [
1389
+ 141,
1390
+ 240,
1391
+ 320,
1392
+ 3
1393
+ ],
1394
+ "dtype": "uint8"
1395
+ },
1396
+ {
1397
+ "path": "obs/sensor_data/top_camera/rgb",
1398
+ "shape": [
1399
+ 141,
1400
+ 320,
1401
+ 240,
1402
+ 3
1403
+ ],
1404
+ "dtype": "uint8"
1405
+ }
1406
+ ],
1407
+ "trajectory_sha256": "7b69a4074c50fa7623cc5b5fe1db0c27dcc2b074467ddeaa85e27d71f280e9e1",
1408
+ "source_sha256": "3e95fa1903eb53e02d80347f9b7dd9c3dd68ae8f6837a353f4dddd62c3b03cb6",
1409
+ "view_sha256": "7bb5938f9629b711d0f28bf1c4a8cf696d568142d96cdec68f62bc55b4a9ace3"
1410
+ },
1411
+ {
1412
+ "demo_id": "demo_13",
1413
+ "file": "demo_13.h5",
1414
+ "length": 140,
1415
+ "fields": [
1416
+ {
1417
+ "path": "actions/panda-0",
1418
+ "shape": [
1419
+ 140,
1420
+ 7
1421
+ ],
1422
+ "dtype": "float32"
1423
+ },
1424
+ {
1425
+ "path": "actions/panda-1",
1426
+ "shape": [
1427
+ 140,
1428
+ 7
1429
+ ],
1430
+ "dtype": "float32"
1431
+ },
1432
+ {
1433
+ "path": "obs/agent/panda-0/qpos",
1434
+ "shape": [
1435
+ 141,
1436
+ 9
1437
+ ],
1438
+ "dtype": "float32"
1439
+ },
1440
+ {
1441
+ "path": "obs/agent/panda-0/qvel",
1442
+ "shape": [
1443
+ 141,
1444
+ 9
1445
+ ],
1446
+ "dtype": "float32"
1447
+ },
1448
+ {
1449
+ "path": "obs/agent/panda-1/qpos",
1450
+ "shape": [
1451
+ 141,
1452
+ 9
1453
+ ],
1454
+ "dtype": "float32"
1455
+ },
1456
+ {
1457
+ "path": "obs/agent/panda-1/qvel",
1458
+ "shape": [
1459
+ 141,
1460
+ 9
1461
+ ],
1462
+ "dtype": "float32"
1463
+ },
1464
+ {
1465
+ "path": "obs/sensor_data/front_camera/rgb",
1466
+ "shape": [
1467
+ 141,
1468
+ 240,
1469
+ 320,
1470
+ 3
1471
+ ],
1472
+ "dtype": "uint8"
1473
+ },
1474
+ {
1475
+ "path": "obs/sensor_data/left_camera/rgb",
1476
+ "shape": [
1477
+ 141,
1478
+ 240,
1479
+ 320,
1480
+ 3
1481
+ ],
1482
+ "dtype": "uint8"
1483
+ },
1484
+ {
1485
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1486
+ "shape": [
1487
+ 141,
1488
+ 240,
1489
+ 320,
1490
+ 3
1491
+ ],
1492
+ "dtype": "uint8"
1493
+ },
1494
+ {
1495
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1496
+ "shape": [
1497
+ 141,
1498
+ 240,
1499
+ 320,
1500
+ 3
1501
+ ],
1502
+ "dtype": "uint8"
1503
+ },
1504
+ {
1505
+ "path": "obs/sensor_data/top_camera/rgb",
1506
+ "shape": [
1507
+ 141,
1508
+ 320,
1509
+ 240,
1510
+ 3
1511
+ ],
1512
+ "dtype": "uint8"
1513
+ }
1514
+ ],
1515
+ "trajectory_sha256": "5d4fbe3382bf0c394a9cd8c4cf07c9c6882b827f792a252a302c082cb1ccf921",
1516
+ "source_sha256": "b1593d563e2d028c6c9449fd751e5ff78e303863ceacd0d97e3c02f13b8870eb",
1517
+ "view_sha256": "c2d9f388a3099e7e725c3cd6a27602985d5c3124155788b79f264c6efbbf5d24"
1518
+ },
1519
+ {
1520
+ "demo_id": "demo_14",
1521
+ "file": "demo_14.h5",
1522
+ "length": 140,
1523
+ "fields": [
1524
+ {
1525
+ "path": "actions/panda-0",
1526
+ "shape": [
1527
+ 140,
1528
+ 7
1529
+ ],
1530
+ "dtype": "float32"
1531
+ },
1532
+ {
1533
+ "path": "actions/panda-1",
1534
+ "shape": [
1535
+ 140,
1536
+ 7
1537
+ ],
1538
+ "dtype": "float32"
1539
+ },
1540
+ {
1541
+ "path": "obs/agent/panda-0/qpos",
1542
+ "shape": [
1543
+ 141,
1544
+ 9
1545
+ ],
1546
+ "dtype": "float32"
1547
+ },
1548
+ {
1549
+ "path": "obs/agent/panda-0/qvel",
1550
+ "shape": [
1551
+ 141,
1552
+ 9
1553
+ ],
1554
+ "dtype": "float32"
1555
+ },
1556
+ {
1557
+ "path": "obs/agent/panda-1/qpos",
1558
+ "shape": [
1559
+ 141,
1560
+ 9
1561
+ ],
1562
+ "dtype": "float32"
1563
+ },
1564
+ {
1565
+ "path": "obs/agent/panda-1/qvel",
1566
+ "shape": [
1567
+ 141,
1568
+ 9
1569
+ ],
1570
+ "dtype": "float32"
1571
+ },
1572
+ {
1573
+ "path": "obs/sensor_data/front_camera/rgb",
1574
+ "shape": [
1575
+ 141,
1576
+ 240,
1577
+ 320,
1578
+ 3
1579
+ ],
1580
+ "dtype": "uint8"
1581
+ },
1582
+ {
1583
+ "path": "obs/sensor_data/left_camera/rgb",
1584
+ "shape": [
1585
+ 141,
1586
+ 240,
1587
+ 320,
1588
+ 3
1589
+ ],
1590
+ "dtype": "uint8"
1591
+ },
1592
+ {
1593
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1594
+ "shape": [
1595
+ 141,
1596
+ 240,
1597
+ 320,
1598
+ 3
1599
+ ],
1600
+ "dtype": "uint8"
1601
+ },
1602
+ {
1603
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1604
+ "shape": [
1605
+ 141,
1606
+ 240,
1607
+ 320,
1608
+ 3
1609
+ ],
1610
+ "dtype": "uint8"
1611
+ },
1612
+ {
1613
+ "path": "obs/sensor_data/top_camera/rgb",
1614
+ "shape": [
1615
+ 141,
1616
+ 320,
1617
+ 240,
1618
+ 3
1619
+ ],
1620
+ "dtype": "uint8"
1621
+ }
1622
+ ],
1623
+ "trajectory_sha256": "9bef23cbae3efade720257a6b20880a5c21f97428bba500500504e888d5b1b8b",
1624
+ "source_sha256": "6c56f135de2673fc322a9cc0b9567677337c7fdb2c4fb6185dbb40794e397357",
1625
+ "view_sha256": "7a4ec477abb581a6f0777b5e3feb8b3355332e22da478f2c36e5e16a922ff9cd"
1626
+ },
1627
+ {
1628
+ "demo_id": "demo_15",
1629
+ "file": "demo_15.h5",
1630
+ "length": 141,
1631
+ "fields": [
1632
+ {
1633
+ "path": "actions/panda-0",
1634
+ "shape": [
1635
+ 141,
1636
+ 7
1637
+ ],
1638
+ "dtype": "float32"
1639
+ },
1640
+ {
1641
+ "path": "actions/panda-1",
1642
+ "shape": [
1643
+ 141,
1644
+ 7
1645
+ ],
1646
+ "dtype": "float32"
1647
+ },
1648
+ {
1649
+ "path": "obs/agent/panda-0/qpos",
1650
+ "shape": [
1651
+ 142,
1652
+ 9
1653
+ ],
1654
+ "dtype": "float32"
1655
+ },
1656
+ {
1657
+ "path": "obs/agent/panda-0/qvel",
1658
+ "shape": [
1659
+ 142,
1660
+ 9
1661
+ ],
1662
+ "dtype": "float32"
1663
+ },
1664
+ {
1665
+ "path": "obs/agent/panda-1/qpos",
1666
+ "shape": [
1667
+ 142,
1668
+ 9
1669
+ ],
1670
+ "dtype": "float32"
1671
+ },
1672
+ {
1673
+ "path": "obs/agent/panda-1/qvel",
1674
+ "shape": [
1675
+ 142,
1676
+ 9
1677
+ ],
1678
+ "dtype": "float32"
1679
+ },
1680
+ {
1681
+ "path": "obs/sensor_data/front_camera/rgb",
1682
+ "shape": [
1683
+ 142,
1684
+ 240,
1685
+ 320,
1686
+ 3
1687
+ ],
1688
+ "dtype": "uint8"
1689
+ },
1690
+ {
1691
+ "path": "obs/sensor_data/left_camera/rgb",
1692
+ "shape": [
1693
+ 142,
1694
+ 240,
1695
+ 320,
1696
+ 3
1697
+ ],
1698
+ "dtype": "uint8"
1699
+ },
1700
+ {
1701
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1702
+ "shape": [
1703
+ 142,
1704
+ 240,
1705
+ 320,
1706
+ 3
1707
+ ],
1708
+ "dtype": "uint8"
1709
+ },
1710
+ {
1711
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1712
+ "shape": [
1713
+ 142,
1714
+ 240,
1715
+ 320,
1716
+ 3
1717
+ ],
1718
+ "dtype": "uint8"
1719
+ },
1720
+ {
1721
+ "path": "obs/sensor_data/top_camera/rgb",
1722
+ "shape": [
1723
+ 142,
1724
+ 320,
1725
+ 240,
1726
+ 3
1727
+ ],
1728
+ "dtype": "uint8"
1729
+ }
1730
+ ],
1731
+ "trajectory_sha256": "7c27ea63098c3b18436c2c3ab81302d6ed6cd41b718f4071c985442cedb529da",
1732
+ "source_sha256": "dc7deb5a5e5bf1e5938d6fc27663e97d9c3fbd69541bc1cbb53afc4ffd4f5e98",
1733
+ "view_sha256": "b627d3db21add58e0eeb7f3b31f7d3cba9629f0a16f7da8dd30c4599b2b84c45"
1734
+ },
1735
+ {
1736
+ "demo_id": "demo_16",
1737
+ "file": "demo_16.h5",
1738
+ "length": 140,
1739
+ "fields": [
1740
+ {
1741
+ "path": "actions/panda-0",
1742
+ "shape": [
1743
+ 140,
1744
+ 7
1745
+ ],
1746
+ "dtype": "float32"
1747
+ },
1748
+ {
1749
+ "path": "actions/panda-1",
1750
+ "shape": [
1751
+ 140,
1752
+ 7
1753
+ ],
1754
+ "dtype": "float32"
1755
+ },
1756
+ {
1757
+ "path": "obs/agent/panda-0/qpos",
1758
+ "shape": [
1759
+ 141,
1760
+ 9
1761
+ ],
1762
+ "dtype": "float32"
1763
+ },
1764
+ {
1765
+ "path": "obs/agent/panda-0/qvel",
1766
+ "shape": [
1767
+ 141,
1768
+ 9
1769
+ ],
1770
+ "dtype": "float32"
1771
+ },
1772
+ {
1773
+ "path": "obs/agent/panda-1/qpos",
1774
+ "shape": [
1775
+ 141,
1776
+ 9
1777
+ ],
1778
+ "dtype": "float32"
1779
+ },
1780
+ {
1781
+ "path": "obs/agent/panda-1/qvel",
1782
+ "shape": [
1783
+ 141,
1784
+ 9
1785
+ ],
1786
+ "dtype": "float32"
1787
+ },
1788
+ {
1789
+ "path": "obs/sensor_data/front_camera/rgb",
1790
+ "shape": [
1791
+ 141,
1792
+ 240,
1793
+ 320,
1794
+ 3
1795
+ ],
1796
+ "dtype": "uint8"
1797
+ },
1798
+ {
1799
+ "path": "obs/sensor_data/left_camera/rgb",
1800
+ "shape": [
1801
+ 141,
1802
+ 240,
1803
+ 320,
1804
+ 3
1805
+ ],
1806
+ "dtype": "uint8"
1807
+ },
1808
+ {
1809
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1810
+ "shape": [
1811
+ 141,
1812
+ 240,
1813
+ 320,
1814
+ 3
1815
+ ],
1816
+ "dtype": "uint8"
1817
+ },
1818
+ {
1819
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1820
+ "shape": [
1821
+ 141,
1822
+ 240,
1823
+ 320,
1824
+ 3
1825
+ ],
1826
+ "dtype": "uint8"
1827
+ },
1828
+ {
1829
+ "path": "obs/sensor_data/top_camera/rgb",
1830
+ "shape": [
1831
+ 141,
1832
+ 320,
1833
+ 240,
1834
+ 3
1835
+ ],
1836
+ "dtype": "uint8"
1837
+ }
1838
+ ],
1839
+ "trajectory_sha256": "0f04250cc83d51753e72c0d8019c330c299c46929435f5b5da71fcec5f631753",
1840
+ "source_sha256": "57dfb7e98970637a9aa8657e807547282da470e5e3b0e51835580b6fac9ff430",
1841
+ "view_sha256": "d89048da08dea2cfdfee7a901ea072ecfe2405befd8798388366d4f69be52731"
1842
+ },
1843
+ {
1844
+ "demo_id": "demo_17",
1845
+ "file": "demo_17.h5",
1846
+ "length": 139,
1847
+ "fields": [
1848
+ {
1849
+ "path": "actions/panda-0",
1850
+ "shape": [
1851
+ 139,
1852
+ 7
1853
+ ],
1854
+ "dtype": "float32"
1855
+ },
1856
+ {
1857
+ "path": "actions/panda-1",
1858
+ "shape": [
1859
+ 139,
1860
+ 7
1861
+ ],
1862
+ "dtype": "float32"
1863
+ },
1864
+ {
1865
+ "path": "obs/agent/panda-0/qpos",
1866
+ "shape": [
1867
+ 140,
1868
+ 9
1869
+ ],
1870
+ "dtype": "float32"
1871
+ },
1872
+ {
1873
+ "path": "obs/agent/panda-0/qvel",
1874
+ "shape": [
1875
+ 140,
1876
+ 9
1877
+ ],
1878
+ "dtype": "float32"
1879
+ },
1880
+ {
1881
+ "path": "obs/agent/panda-1/qpos",
1882
+ "shape": [
1883
+ 140,
1884
+ 9
1885
+ ],
1886
+ "dtype": "float32"
1887
+ },
1888
+ {
1889
+ "path": "obs/agent/panda-1/qvel",
1890
+ "shape": [
1891
+ 140,
1892
+ 9
1893
+ ],
1894
+ "dtype": "float32"
1895
+ },
1896
+ {
1897
+ "path": "obs/sensor_data/front_camera/rgb",
1898
+ "shape": [
1899
+ 140,
1900
+ 240,
1901
+ 320,
1902
+ 3
1903
+ ],
1904
+ "dtype": "uint8"
1905
+ },
1906
+ {
1907
+ "path": "obs/sensor_data/left_camera/rgb",
1908
+ "shape": [
1909
+ 140,
1910
+ 240,
1911
+ 320,
1912
+ 3
1913
+ ],
1914
+ "dtype": "uint8"
1915
+ },
1916
+ {
1917
+ "path": "obs/sensor_data/robot_camera_0/rgb",
1918
+ "shape": [
1919
+ 140,
1920
+ 240,
1921
+ 320,
1922
+ 3
1923
+ ],
1924
+ "dtype": "uint8"
1925
+ },
1926
+ {
1927
+ "path": "obs/sensor_data/robot_camera_1/rgb",
1928
+ "shape": [
1929
+ 140,
1930
+ 240,
1931
+ 320,
1932
+ 3
1933
+ ],
1934
+ "dtype": "uint8"
1935
+ },
1936
+ {
1937
+ "path": "obs/sensor_data/top_camera/rgb",
1938
+ "shape": [
1939
+ 140,
1940
+ 320,
1941
+ 240,
1942
+ 3
1943
+ ],
1944
+ "dtype": "uint8"
1945
+ }
1946
+ ],
1947
+ "trajectory_sha256": "7c357078603cd75b8c69ebc05c5e0b5612156f5c36e888bae4ed0a5bb88aa1ed",
1948
+ "source_sha256": "3ba7d2426158186d0fff6e86051e55598acfb138848289872256a9535120c96e",
1949
+ "view_sha256": "cae1a156aa92f9c18f3f6279915f48832e95d36f473cf3292e0020f4abdcbca2"
1950
+ },
1951
+ {
1952
+ "demo_id": "demo_18",
1953
+ "file": "demo_18.h5",
1954
+ "length": 139,
1955
+ "fields": [
1956
+ {
1957
+ "path": "actions/panda-0",
1958
+ "shape": [
1959
+ 139,
1960
+ 7
1961
+ ],
1962
+ "dtype": "float32"
1963
+ },
1964
+ {
1965
+ "path": "actions/panda-1",
1966
+ "shape": [
1967
+ 139,
1968
+ 7
1969
+ ],
1970
+ "dtype": "float32"
1971
+ },
1972
+ {
1973
+ "path": "obs/agent/panda-0/qpos",
1974
+ "shape": [
1975
+ 140,
1976
+ 9
1977
+ ],
1978
+ "dtype": "float32"
1979
+ },
1980
+ {
1981
+ "path": "obs/agent/panda-0/qvel",
1982
+ "shape": [
1983
+ 140,
1984
+ 9
1985
+ ],
1986
+ "dtype": "float32"
1987
+ },
1988
+ {
1989
+ "path": "obs/agent/panda-1/qpos",
1990
+ "shape": [
1991
+ 140,
1992
+ 9
1993
+ ],
1994
+ "dtype": "float32"
1995
+ },
1996
+ {
1997
+ "path": "obs/agent/panda-1/qvel",
1998
+ "shape": [
1999
+ 140,
2000
+ 9
2001
+ ],
2002
+ "dtype": "float32"
2003
+ },
2004
+ {
2005
+ "path": "obs/sensor_data/front_camera/rgb",
2006
+ "shape": [
2007
+ 140,
2008
+ 240,
2009
+ 320,
2010
+ 3
2011
+ ],
2012
+ "dtype": "uint8"
2013
+ },
2014
+ {
2015
+ "path": "obs/sensor_data/left_camera/rgb",
2016
+ "shape": [
2017
+ 140,
2018
+ 240,
2019
+ 320,
2020
+ 3
2021
+ ],
2022
+ "dtype": "uint8"
2023
+ },
2024
+ {
2025
+ "path": "obs/sensor_data/robot_camera_0/rgb",
2026
+ "shape": [
2027
+ 140,
2028
+ 240,
2029
+ 320,
2030
+ 3
2031
+ ],
2032
+ "dtype": "uint8"
2033
+ },
2034
+ {
2035
+ "path": "obs/sensor_data/robot_camera_1/rgb",
2036
+ "shape": [
2037
+ 140,
2038
+ 240,
2039
+ 320,
2040
+ 3
2041
+ ],
2042
+ "dtype": "uint8"
2043
+ },
2044
+ {
2045
+ "path": "obs/sensor_data/top_camera/rgb",
2046
+ "shape": [
2047
+ 140,
2048
+ 320,
2049
+ 240,
2050
+ 3
2051
+ ],
2052
+ "dtype": "uint8"
2053
+ }
2054
+ ],
2055
+ "trajectory_sha256": "9442a3256900b7c95e63d6b10f2ba17eb4777b5250535f053bdf2258c1c1d7e9",
2056
+ "source_sha256": "b56c9cb5f7bc8564a98a825103a0285b9947a92e52f115a3cc5a2ed255efbdd1",
2057
+ "view_sha256": "7d22f6ca17154749a728474ebb4ce19c3bc3b4cb0b3f5774c9be53042c954cb6"
2058
+ },
2059
+ {
2060
+ "demo_id": "demo_19",
2061
+ "file": "demo_19.h5",
2062
+ "length": 140,
2063
+ "fields": [
2064
+ {
2065
+ "path": "actions/panda-0",
2066
+ "shape": [
2067
+ 140,
2068
+ 7
2069
+ ],
2070
+ "dtype": "float32"
2071
+ },
2072
+ {
2073
+ "path": "actions/panda-1",
2074
+ "shape": [
2075
+ 140,
2076
+ 7
2077
+ ],
2078
+ "dtype": "float32"
2079
+ },
2080
+ {
2081
+ "path": "obs/agent/panda-0/qpos",
2082
+ "shape": [
2083
+ 141,
2084
+ 9
2085
+ ],
2086
+ "dtype": "float32"
2087
+ },
2088
+ {
2089
+ "path": "obs/agent/panda-0/qvel",
2090
+ "shape": [
2091
+ 141,
2092
+ 9
2093
+ ],
2094
+ "dtype": "float32"
2095
+ },
2096
+ {
2097
+ "path": "obs/agent/panda-1/qpos",
2098
+ "shape": [
2099
+ 141,
2100
+ 9
2101
+ ],
2102
+ "dtype": "float32"
2103
+ },
2104
+ {
2105
+ "path": "obs/agent/panda-1/qvel",
2106
+ "shape": [
2107
+ 141,
2108
+ 9
2109
+ ],
2110
+ "dtype": "float32"
2111
+ },
2112
+ {
2113
+ "path": "obs/sensor_data/front_camera/rgb",
2114
+ "shape": [
2115
+ 141,
2116
+ 240,
2117
+ 320,
2118
+ 3
2119
+ ],
2120
+ "dtype": "uint8"
2121
+ },
2122
+ {
2123
+ "path": "obs/sensor_data/left_camera/rgb",
2124
+ "shape": [
2125
+ 141,
2126
+ 240,
2127
+ 320,
2128
+ 3
2129
+ ],
2130
+ "dtype": "uint8"
2131
+ },
2132
+ {
2133
+ "path": "obs/sensor_data/robot_camera_0/rgb",
2134
+ "shape": [
2135
+ 141,
2136
+ 240,
2137
+ 320,
2138
+ 3
2139
+ ],
2140
+ "dtype": "uint8"
2141
+ },
2142
+ {
2143
+ "path": "obs/sensor_data/robot_camera_1/rgb",
2144
+ "shape": [
2145
+ 141,
2146
+ 240,
2147
+ 320,
2148
+ 3
2149
+ ],
2150
+ "dtype": "uint8"
2151
+ },
2152
+ {
2153
+ "path": "obs/sensor_data/top_camera/rgb",
2154
+ "shape": [
2155
+ 141,
2156
+ 320,
2157
+ 240,
2158
+ 3
2159
+ ],
2160
+ "dtype": "uint8"
2161
+ }
2162
+ ],
2163
+ "trajectory_sha256": "9709de74f190b62959e819cd981b07884971b125a95f208266f612982c97378c",
2164
+ "source_sha256": "e59ccf1719ce4b1cadcc6544e84c67894be24686280ab82d021fcb601b15bd61",
2165
+ "view_sha256": "68c1944c9c7c252d7fb4b3707491eb59df590f5add2510e3103100e18c345e50"
2166
+ }
2167
+ ]
2168
+ }
agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/environment.lock ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ absl-py==2.4.0
2
+ accelerate==1.13.0
3
+ addict==2.4.0
4
+ aiohappyeyeballs==2.6.1
5
+ aiohttp==3.13.5
6
+ aiosignal==1.4.0
7
+ annotated-types==0.7.0
8
+ anyio==4.13.0
9
+ anytree==2.13.0
10
+ asttokens==3.0.1
11
+ async-timeout==5.0.1
12
+ attrs==26.1.0
13
+ av==15.1.0
14
+ blinker==1.9.0
15
+ casadi==3.7.2
16
+ certifi==2026.4.22
17
+ cffi==2.0.0
18
+ charset-normalizer==3.4.7
19
+ click==8.3.3
20
+ cloudpickle==3.1.2
21
+ cmake==4.1.3
22
+ coacd==1.0.11
23
+ colorama==0.4.6
24
+ colorlog==6.10.1
25
+ comm==0.2.3
26
+ ConfigArgParse==1.7.5
27
+ contourpy==1.3.2
28
+ cuda-bindings==12.9.4
29
+ cuda-pathfinder==1.5.4
30
+ cycler==0.12.1
31
+ dacite==1.9.2
32
+ dash==4.1.0
33
+ dataclasses-json==0.6.7
34
+ datasets==4.8.5
35
+ decorator==5.2.1
36
+ deepdiff==8.6.2
37
+ Deprecated==1.3.1
38
+ diffusers==0.35.2
39
+ dill==0.4.1
40
+ docstring_parser==0.18.0
41
+ draccus==0.10.0
42
+ einops==0.8.2
43
+ embreex==4.4.0
44
+ etils==1.13.0
45
+ evdev==1.9.3
46
+ exceptiongroup==1.3.1
47
+ executing==2.2.1
48
+ Farama-Notifications==0.0.6
49
+ fast_simplification==0.1.13
50
+ fastjsonschema==2.21.2
51
+ filelock==3.29.0
52
+ Flask==3.1.3
53
+ fonttools==4.63.0
54
+ freetype-py==2.5.1
55
+ frozenlist==1.8.0
56
+ fsspec==2026.2.0
57
+ genesis-world @ file:///data/gck/genesis-world
58
+ git-filter-repo==2.47.0
59
+ gitdb==4.0.12
60
+ GitPython==3.1.50
61
+ glfw==2.10.0
62
+ gs-madrona==0.0.5
63
+ gstaichi==3.3.0
64
+ gym==0.26.2
65
+ gym-notices==0.1.0
66
+ gymnasium==1.3.0
67
+ h11==0.16.0
68
+ h5py==3.16.0
69
+ hf-xet==1.5.0
70
+ hf_transfer==0.1.9
71
+ httpcore==1.0.9
72
+ httpx==0.28.1
73
+ huggingface-hub==0.35.3
74
+ idna==3.15
75
+ ImageIO==2.37.3
76
+ imageio-ffmpeg==0.6.0
77
+ importlib_metadata==9.0.0
78
+ importlib_resources==7.1.0
79
+ inquirerpy==0.3.4
80
+ ipython==8.39.0
81
+ ipywidgets==8.1.8
82
+ itsdangerous==2.2.0
83
+ jedi==0.20.0
84
+ Jinja2==3.1.6
85
+ joblib==1.5.3
86
+ jsonlines==4.0.0
87
+ jsonschema==4.26.0
88
+ jsonschema-specifications==2025.9.1
89
+ jupyter_core==5.9.1
90
+ jupyterlab_widgets==3.0.16
91
+ kiwisolver==1.5.0
92
+ lazy-loader==0.5
93
+ lerobot==0.4.4
94
+ libigl==2.6.2
95
+ llvmlite==0.47.0
96
+ loguru==0.7.3
97
+ lxml==6.1.0
98
+ manifold3d==3.4.1
99
+ -e git+ssh://git@github.com/HeegerGao/maniladder.git@f6eed17496816bb41ce484f2d7580990a908d3e6#egg=maniladder
100
+ mapbox_earcut==2.0.0
101
+ markdown-it-py==4.2.0
102
+ MarkupSafe==3.0.3
103
+ marshmallow==3.26.2
104
+ matplotlib==3.10.9
105
+ matplotlib-inline==0.2.2
106
+ mdurl==0.1.2
107
+ mergedeep==1.3.4
108
+ moviepy==2.2.1
109
+ mpmath==1.3.0
110
+ msgspec==0.21.1
111
+ mujoco==3.8.1
112
+ multidict==6.7.1
113
+ multiprocess==0.70.19
114
+ mypy_extensions==1.1.0
115
+ nbformat==5.10.4
116
+ nest-asyncio==1.6.0
117
+ networkx==3.4.2
118
+ ninja==1.13.0
119
+ nlopt==2.10.0
120
+ numba==0.65.1
121
+ numpy @ file:///home/conda/feedstock_root/build_artifacts/numpy_1747544640217/work/dist/numpy-2.2.6-cp310-cp310-linux_x86_64.whl#sha256=d6d964caeef85d00073d27cd62b46883d275b3d8162f723f0fcabbd0b3cc3f9d
122
+ nvidia-cublas-cu12==12.8.4.1
123
+ nvidia-cuda-cupti-cu12==12.8.90
124
+ nvidia-cuda-nvrtc-cu12==12.8.93
125
+ nvidia-cuda-runtime-cu12==12.8.90
126
+ nvidia-cudnn-cu12==9.10.2.21
127
+ nvidia-cufft-cu12==11.3.3.83
128
+ nvidia-cufile-cu12==1.13.1.3
129
+ nvidia-curand-cu12==10.3.9.90
130
+ nvidia-cusolver-cu12==11.7.3.90
131
+ nvidia-cusparse-cu12==12.5.8.93
132
+ nvidia-cusparselt-cu12==0.7.1
133
+ nvidia-nccl-cu12==2.27.5
134
+ nvidia-nvjitlink-cu12==12.8.93
135
+ nvidia-nvshmem-cu12==3.4.5
136
+ nvidia-nvtx-cu12==12.8.90
137
+ open3d==0.19.0
138
+ opencv-python==4.13.0.92
139
+ opencv-python-headless==4.12.0.88
140
+ OpenEXR==3.4.11
141
+ orderly-set==5.5.0
142
+ packaging==25.0
143
+ pandas==2.3.3
144
+ parso==0.8.7
145
+ pexpect==4.9.0
146
+ pfzy==0.3.4
147
+ pillow==11.3.0
148
+ pin==4.0.0
149
+ pip @ file:///home/task_177498883471825/croot/pip_1774988882575/work
150
+ platformdirs==4.9.6
151
+ plotly==5.24.1
152
+ polyscope==2.6.1
153
+ proglog==0.1.12
154
+ prompt_toolkit==3.0.52
155
+ propcache==0.5.2
156
+ protobuf==6.33.6
157
+ psutil==7.2.2
158
+ ptyprocess==0.7.0
159
+ pure_eval==0.2.3
160
+ py-cpuinfo==9.0.0
161
+ py-spy==0.4.2
162
+ pyarrow==24.0.0
163
+ pycollada==0.9.3
164
+ pycparser==3.0
165
+ pydantic==2.13.4
166
+ pydantic_core==2.46.4
167
+ PyGEL3D==0.6.1
168
+ pyglet==2.1.14
169
+ pygltflib==1.16.0
170
+ Pygments==2.20.0
171
+ pymeshlab==2025.7.post1
172
+ pynput==1.8.2
173
+ PyOpenGL==3.1.10
174
+ pyparsing==3.3.2
175
+ pyperclip==1.11.0
176
+ PyQt6==6.11.0
177
+ PyQt6-Qt6==6.11.0
178
+ PyQt6_sip==13.11.1
179
+ pyquaternion==0.9.9
180
+ pyserial==3.5
181
+ pysplashsurf==0.12.0.2
182
+ python-dateutil==2.9.0.post0
183
+ python-dotenv==1.2.2
184
+ python-xlib==0.33
185
+ pytransform3d==3.15.0
186
+ pytz==2026.2
187
+ pyuipc @ file:///tmp/claude-1005/-mnt4-gck-maniladder/9f005e67-04f6-4368-bd0e-9f40896c7dbf/scratchpad/pyuipc_dl/pyuipc-0.0.25-cp310-cp310-manylinux_2_34_x86_64.whl#sha256=76ff13554e1971ea4401db2c875369cf32352ae3a5a4c398c45193231b5cbf00
188
+ PyYAML==6.0.3
189
+ pyyaml-include==1.4.1
190
+ referencing==0.37.0
191
+ regex==2026.5.9
192
+ requests==2.34.2
193
+ rerun-sdk==0.26.2
194
+ retrying==1.4.2
195
+ rich==14.3.4
196
+ rpds-py==0.30.0
197
+ rtree==1.4.1
198
+ safetensors==0.7.0
199
+ sapien==3.0.3
200
+ scikit-image==0.25.2
201
+ scikit-learn==1.7.2
202
+ scipy @ file:///home/conda/feedstock_root/build_artifacts/scipy-split_1739790642651/work/dist/scipy-1.15.2-cp310-cp310-linux_x86_64.whl#sha256=9e52bad6c3294d1a5b04a13632118ca2157130603c6c018c2d710162b223b27e
203
+ sentry-sdk==2.60.0
204
+ setuptools==80.10.2
205
+ shapely==2.1.2
206
+ six==1.17.0
207
+ smmap==5.0.3
208
+ stack-data==0.6.3
209
+ svg.path==7.0
210
+ sympy==1.14.0
211
+ tenacity==9.1.4
212
+ termcolor==3.3.0
213
+ tetgen==0.8.4
214
+ threadpoolctl==3.6.0
215
+ tifffile==2025.5.10
216
+ timm==1.0.27
217
+ toml==0.10.2
218
+ torch==2.10.0
219
+ torchaudio==2.11.0
220
+ torchcodec==0.10.0
221
+ torchvision==0.25.0
222
+ tqdm==4.67.3
223
+ traitlets==5.15.0
224
+ transforms3d==0.4.2
225
+ trimesh==4.12.2
226
+ triton==3.6.0
227
+ typeguard==4.5.2
228
+ typing-inspect==0.9.0
229
+ typing-inspection==0.4.2
230
+ typing_extensions==4.15.0
231
+ tyro==1.0.13
232
+ tzdata==2026.2
233
+ urllib3==2.7.0
234
+ vhacdx==0.0.10
235
+ viser==1.0.27
236
+ vtk==9.6.1
237
+ wandb==0.24.2
238
+ wcwidth==0.7.0
239
+ websockets==16.0
240
+ Werkzeug==3.1.8
241
+ wheel==0.46.3
242
+ widgetsnbextension==4.0.15
243
+ wrapt==2.1.2
244
+ xxhash==3.7.0
245
+ yarl==1.23.0
246
+ yourdfpy==0.0.60
247
+ z3-solver==4.16.0.0
248
+ zipp==3.23.1
249
+ zstandard==0.25.0
agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/freeze_manifest.json ADDED
@@ -0,0 +1,117 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "files": {
3
+ "ball_reference.npy": "2713eb52580b4952546307c259b40cabe0030fb307b15afb192cc49a43ccbf20",
4
+ "controller.py": "2b8eab423812f8158fac9bb1c92e82adf300fbc743868579875eca3620118b7e",
5
+ "demo_manifest.json": "1fdf1884267a8feb6a4d8c064554f82606848cebf5020ad66ed4be01807bc1e2",
6
+ "demonstrations_readonly/demo_00.h5": "fc2accfe6d35fa8d7395d00a18969504b4157d7aa67aca75d64baa98b06e57a6",
7
+ "demonstrations_readonly/demo_01.h5": "564a89cf86c685206ab6442adaee4b7d3a67e59805c97bc37d0d8a2a9e43f8ec",
8
+ "demonstrations_readonly/demo_02.h5": "d3109687e72305fa5f6b97bf4805dc2260072d0197018c9299d9344282e872c1",
9
+ "demonstrations_readonly/demo_03.h5": "44e8f69c7f533044cc782e9b69b74f4f2fe0c6bca2ef2cd2a987f263125e4a22",
10
+ "demonstrations_readonly/demo_04.h5": "5f7505253cbdbfcb343f05d4a1a0ce694ba54e28a43e6b361af32a9f29f99d17",
11
+ "demonstrations_readonly/demo_05.h5": "e90836977e0cdae992916d6f25e8be7e9093bf20f14df45fc10ad9bf66437b24",
12
+ "demonstrations_readonly/demo_06.h5": "da208e25be7a82564362249ad8247b88acb9321dba9b179ff0a99c830b14b96d",
13
+ "demonstrations_readonly/demo_07.h5": "88b6bcf7a83f9fbf473820b41f71f1df19279efe9f0ed46a5f78ba5c93928f9e",
14
+ "demonstrations_readonly/demo_08.h5": "f25ee42499e3548463aab47927281d5199527dc19dd4ea5f134f6f583164619b",
15
+ "demonstrations_readonly/demo_09.h5": "fa3eb0447d6b9741bce6057773220476551678883346bc5cd2a438b6b0deac40",
16
+ "demonstrations_readonly/demo_10.h5": "7eb12dd94e5c0b88e933fa049293bd77929845f044d6d73766f6a3ffcbe950ae",
17
+ "demonstrations_readonly/demo_11.h5": "b8a82cb74ab3edc42993c206c3994936fb53697a9ab685659bf5c790b83a9913",
18
+ "demonstrations_readonly/demo_12.h5": "7bb5938f9629b711d0f28bf1c4a8cf696d568142d96cdec68f62bc55b4a9ace3",
19
+ "demonstrations_readonly/demo_13.h5": "c2d9f388a3099e7e725c3cd6a27602985d5c3124155788b79f264c6efbbf5d24",
20
+ "demonstrations_readonly/demo_14.h5": "7a4ec477abb581a6f0777b5e3feb8b3355332e22da478f2c36e5e16a922ff9cd",
21
+ "demonstrations_readonly/demo_15.h5": "b627d3db21add58e0eeb7f3b31f7d3cba9629f0a16f7da8dd30c4599b2b84c45",
22
+ "demonstrations_readonly/demo_16.h5": "d89048da08dea2cfdfee7a901ea072ecfe2405befd8798388366d4f69be52731",
23
+ "demonstrations_readonly/demo_17.h5": "cae1a156aa92f9c18f3f6279915f48832e95d36f473cf3292e0020f4abdcbca2",
24
+ "demonstrations_readonly/demo_18.h5": "7d22f6ca17154749a728474ebb4ce19c3bc3b4cb0b3f5774c9be53042c954cb6",
25
+ "demonstrations_readonly/demo_19.h5": "68c1944c9c7c252d7fb4b3707491eb59df590f5add2510e3103100e18c345e50",
26
+ "environment.lock": "f9f11f093744865020243315494d48c69e38ff9e67783e7cf5f0a1e5f22b34aa",
27
+ "findings.md": "dcf65821dc9b1b101c5de2464dec6c5c3ee20a3fd79277890eb50930ca5c5267",
28
+ "handoff.md": "a9660d801710fcc6fbe744470a10cf1813aa881b3703bfc97ea85291960510d0",
29
+ "kinematics.py": "66b5bbb372f5a6ec9dd036e0fa3783002182452afa53603b933e7c0dfbcf41ab",
30
+ "model_tool_manifest.yaml": "ff054f46f77d4baec8baee7b376cf889d3411c518d0e5406c021d91fd00bad3b",
31
+ "policy_runner.py": "cd8a30fc763583fb9bf6570414f01d56aba16241c733fc0cdfdfab0f8f27270e",
32
+ "reference.npz": "f06a0bbd10f96cc866fd693b3db2057b9ca34c616d557739df24aa37a80e8b70",
33
+ "skills.md": "e5b840087f04d63110e44c118f28e4e5f8333e2a7d7c8aa67285abcb65450903",
34
+ "vision.py": "133e498054618de50328a4b89d27cbf5ab8d5474d5b29d6383fb1cdd1b8b6b48",
35
+ "worker_api.py": "88cca5fa5c1c9570ca2bd90e8848f67c09224fa498d94e63634ad191e913c41e"
36
+ },
37
+ "bundle_hash": "cffcc9777268838ff107bdfcd455c3e67d820c22adb46f827b87ba0edb2a37d2",
38
+ "exploration_ledger_hash": "dc2178b7bbc3c560cf5039d214954efac7474aef46aa52b6422fcf74d51963d2",
39
+ "entrypoint": "controller.reset_episode(); controller.act(obs)",
40
+ "selection": {
41
+ "rule": "exact-version success rate, valid rehearsal count, recency",
42
+ "rehearsals": [
43
+ {
44
+ "episode": 3,
45
+ "bundle_hash": "bbe8da3d4a644f0ad82c885f057116fb5fe9ce91e88b7bba143dd706a1933512",
46
+ "independent": true,
47
+ "valid": true,
48
+ "status": "finished",
49
+ "success": false,
50
+ "score_source": "native_step_feedback"
51
+ },
52
+ {
53
+ "episode": 10,
54
+ "bundle_hash": "c07bad9cd66b8bd1b7c0671e48211a25872fa76ca5c6081ecdcd052c439a2239",
55
+ "independent": true,
56
+ "valid": true,
57
+ "status": "finished",
58
+ "success": false,
59
+ "score_source": "native_step_feedback"
60
+ },
61
+ {
62
+ "episode": 15,
63
+ "bundle_hash": "cffcc9777268838ff107bdfcd455c3e67d820c22adb46f827b87ba0edb2a37d2",
64
+ "independent": true,
65
+ "valid": true,
66
+ "status": "finished",
67
+ "success": false,
68
+ "score_source": "native_step_feedback"
69
+ },
70
+ {
71
+ "episode": 16,
72
+ "bundle_hash": "cffcc9777268838ff107bdfcd455c3e67d820c22adb46f827b87ba0edb2a37d2",
73
+ "independent": true,
74
+ "valid": true,
75
+ "status": "finished",
76
+ "success": false,
77
+ "score_source": "native_step_feedback"
78
+ },
79
+ {
80
+ "episode": 17,
81
+ "bundle_hash": "cffcc9777268838ff107bdfcd455c3e67d820c22adb46f827b87ba0edb2a37d2",
82
+ "independent": true,
83
+ "valid": true,
84
+ "status": "finished",
85
+ "success": false,
86
+ "score_source": "native_step_feedback"
87
+ },
88
+ {
89
+ "episode": 18,
90
+ "bundle_hash": "cffcc9777268838ff107bdfcd455c3e67d820c22adb46f827b87ba0edb2a37d2",
91
+ "independent": true,
92
+ "valid": true,
93
+ "status": "finished",
94
+ "success": false,
95
+ "score_source": "native_step_feedback"
96
+ },
97
+ {
98
+ "episode": 19,
99
+ "bundle_hash": "cffcc9777268838ff107bdfcd455c3e67d820c22adb46f827b87ba0edb2a37d2",
100
+ "independent": true,
101
+ "valid": true,
102
+ "status": "finished",
103
+ "success": false,
104
+ "score_source": "native_step_feedback"
105
+ },
106
+ {
107
+ "episode": 20,
108
+ "bundle_hash": "cffcc9777268838ff107bdfcd455c3e67d820c22adb46f827b87ba0edb2a37d2",
109
+ "independent": true,
110
+ "valid": true,
111
+ "status": "finished",
112
+ "success": false,
113
+ "score_source": "native_step_feedback"
114
+ }
115
+ ]
116
+ }
117
+ }
agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/handoff.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Runtime and audit handoff
2
+
3
+ controller.py exports reset_episode() and act(obs). Runtime dependencies: Python3, NumPy, OpenCV; local files controller.py, vision.py, kinematics.py, reference.npz, ball_reference.npy. No internet, models, downloads, worker API, task simulator, or development agent is required by act().
4
+
5
+ Native horizon300 at20Hz; initial runner hold consumes step1. This candidate returns None at step210 unless the runner terminates earlier. Action order panda-0 then panda-1, each xyz, rotationxyz, gripper; flat14 values within native [-1,1]. Reset controller memory per episode.
6
+
7
+ This runtime behavior achieved native success in interactive episode9 and failed independent episode10. Episodes15–20 are reserved for unchanged exact-bundle independent validation; their authoritative results and hashes are recorded outside the bundle by the trusted API. No prospective success rate is claimed here. Freeze selects the eligible exact version under the prescribed ranking.
8
+
9
+ Audit evidence remains in /workspace/demo_analysis.md, /workspace/evidence, and /workspace/exploration. Each interactive trial has saved observations; most have full action traces. Episode9's action trace was lost on an interactive harness exception after native termination; its saved native success return and images are preserved, and the limitation is documented. Candidate versions and failures are retained. Independent execution is always entrusted to API.rehearse().
10
+
11
+ The runner adds the verified demo manifest, readonly demo files, and common environment lock. The bundled builder and old synthetic check have been removed from runtime packaging; their originals remain in archived candidate versions. Analytical kinematics is a pure coordinate calculation from qpos, with no simulated dynamics or environment construction.
agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/kinematics.py ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ import numpy as np
2
+
3
+ def fk(q):
4
+ T=np.eye(4)
5
+ for th,a,d,al in zip(q[:7],[0,0,0,.0825,-.0825,0,.088],[.333,0,.316,0,.384,0,0],[0,-np.pi/2,np.pi/2,np.pi/2,-np.pi/2,np.pi/2,np.pi/2]):
6
+ c,s=np.cos(th),np.sin(th);ca,sa=np.cos(al),np.sin(al)
7
+ T=T@np.array([[c,-s,0,a],[s*ca,c*ca,-sa,-d*sa],[s*sa,c*sa,ca,d*ca],[0,0,0,1.]])
8
+ return (T@np.array([0,0,.2104,1]))[:3]
agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/model_tool_manifest.yaml ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "stage": "FINAL EXACT-BUNDLE INDEPENDENT REHEARSALS",
3
+ "status": "ready_for_episodes_15_to_20",
4
+ "task_id": "DualArmLiftBall-v1",
5
+ "neural_finetuning": false,
6
+ "pretrained_weights_modified": false,
7
+ "frozen_weight_references": [],
8
+ "coding_model": {
9
+ "name": "Codex (system-provided)",
10
+ "exact_weight_id": "not exposed",
11
+ "modified": false
12
+ },
13
+ "runtime_dependencies": {
14
+ "python": "3.10.20",
15
+ "numpy": "2.2.6",
16
+ "opencv": "4.12.0"
17
+ },
18
+ "preparation_dependencies": {
19
+ "h5py": "3.16.0",
20
+ "scipy": "1.15.2"
21
+ },
22
+ "tools_used": [
23
+ "functions.exec / exec_command: local shell and Python",
24
+ "view_image: six generated contact sheets",
25
+ "worker_api.API.read_demo: authorized RGB/proprioception/actions",
26
+ "Python standard library: importlib, json, pathlib, hashlib, sys, platform",
27
+ "NumPy: statistics and nonparametric reference arrays",
28
+ "OpenCV: RGB resize, sheet annotation and image writing",
29
+ "h5py: read_demo backend",
30
+ "SciPy: version inspection only",
31
+ "Trusted worker API: budget_status, next_episode, execute_actions, latest_observation, finish_episode, rehearse",
32
+ "Analytical serial-chain kinematics from legal qpos; no dynamics simulator",
33
+ "OpenCV HoughCircles and handcrafted feedback"
34
+ ],
35
+ "provenance": {
36
+ "contract_sha256": "2b28c539c6645273991e1a59f6c8997392ea7d8e3154af09e65af6297154f927",
37
+ "manifest_sha256": "18afc20b05b97f76b130f6c49083fbec231f0b196b087ca6517ccb83115a6ded",
38
+ "worker_api_sha256": "88cca5fa5c1c9570ca2bd90e8848f67c09224fa498d94e63634ad191e913c41e",
39
+ "demo_ids": [
40
+ "demo_00",
41
+ "demo_01",
42
+ "demo_02",
43
+ "demo_03",
44
+ "demo_04",
45
+ "demo_05",
46
+ "demo_06",
47
+ "demo_07",
48
+ "demo_08",
49
+ "demo_09",
50
+ "demo_10",
51
+ "demo_11",
52
+ "demo_12",
53
+ "demo_13",
54
+ "demo_14",
55
+ "demo_15",
56
+ "demo_16",
57
+ "demo_17",
58
+ "demo_18",
59
+ "demo_19"
60
+ ],
61
+ "reference": "reference.npz: all 2819 legal observations; 2799 expert actions plus 20 synthetic endpoint holds",
62
+ "reference_sha256": "f06a0bbd10f96cc866fd693b3db2057b9ca34c616d557739df24aa37a80e8b70",
63
+ "inspection_log": "../evidence/visual_inspection.json"
64
+ },
65
+ "validation": [
66
+ "Authorized interactive exploration; native success observed in episode9",
67
+ "Independent episode10 runnable, native failure",
68
+ "Final unchanged exact bundle to be independently rehearsed in episodes15-20; outcomes recorded by runner"
69
+ ],
70
+ "limitations": [
71
+ "Robustness across ball offsets unproven",
72
+ "No validated regrasp",
73
+ "Classical vision can lose occluded ball",
74
+ "Episode9 interactive action trace incomplete after harness exception",
75
+ "No official success inferred from visual lift",
76
+ "Verified demo manifest and common environment lock supplied by runner"
77
+ ],
78
+ "exploration_episodes_before_final_rehearsals": 14,
79
+ "runtime_files": [
80
+ "controller.py",
81
+ "vision.py",
82
+ "kinematics.py",
83
+ "reference.npz",
84
+ "ball_reference.npy"
85
+ ],
86
+ "runtime_sha256": {
87
+ "controller.py": "2b8eab423812f8158fac9bb1c92e82adf300fbc743868579875eca3620118b7e",
88
+ "vision.py": "133e498054618de50328a4b89d27cbf5ab8d5474d5b29d6383fb1cdd1b8b6b48",
89
+ "kinematics.py": "66b5bbb372f5a6ec9dd036e0fa3783002182452afa53603b933e7c0dfbcf41ab",
90
+ "reference.npz": "f06a0bbd10f96cc866fd693b3db2057b9ca34c616d557739df24aa37a80e8b70",
91
+ "ball_reference.npy": "2713eb52580b4952546307c259b40cabe0030fb307b15afb192cc49a43ccbf20"
92
+ }
93
+ }
agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/skills.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Compiled controller capabilities and limits
2
+
3
+ Uses exactly this task's D20 reference bank and legal RGB, proprioception and step_id. No neural model, fine-tuning, network access or simulator is used by the controller.
4
+
5
+ Classical top-camera Hough-circle localization normalizes images to 480x640. Full-width search and temporal center/radius checks reduce switches to the support. Missing or occluded detections remain possible. Demo00 supplies nominal bilateral approach, tilt, closure, and lift timing. The reference bank preserves all20 demos; runtime deliberately uses demo00 after initial localization.
6
+
7
+ Measured depth feedback uses analytical serial-chain Panda forward kinematics from qpos. A handcrafted pixel-to-depth calibration from legal demos gives target=.00221695*ball_x-.530522. Root-x actions separately correct each arm's depth error with gain5 and cap.18. This is an approximate geometry-based controller, not a trained neural policy or a simulator.
8
+
9
+ Through phase139 the program follows nominal demonstration channels except corrected depth. Phase140–179 extends lift with z=.1 and opposite rx=.03, and regulates lateral tip spacing toward local-x=.58. It then holds closed grippers and finishes at native step210. Native horizon remains300. All actions are clipped14-component float32 vectors in contract order. Runner owns initial hold.
10
+
11
+ Observed limitations: unstable sphere rolling at larger depth offsets, occlusion, geometric calibration error, sensitivity to wrist configuration, and no validated regrasp. The controller never declares success. Only native API feedback or the prescribed conservative visible rule can establish an outcome.
agent_d20_e20/frozen_20261102/tasks/DualArmLiftBall-v1/frozen/vision.py ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import cv2,numpy as np
2
+ cv2.setNumThreads(1)
3
+ def ball(im,previous=None):
4
+ im=cv2.resize(np.asarray(im).squeeze(),(480,640));gray=cv2.cvtColor(im,cv2.COLOR_RGB2GRAY)
5
+ crop=gray[160:480,:]
6
+ cs=cv2.HoughCircles(crop,cv2.HOUGH_GRADIENT,1,35,param1=100,param2=23,minRadius=60,maxRadius=90)
7
+ if cs is None:return None
8
+ cs=cs[0];cs[:,1]+=160
9
+ target=np.array([240.,320.,67.]) if previous is None else previous
10
+ c=min(cs,key=lambda c:np.linalg.norm(c[:2]-target[:2])+2*abs(c[2]-target[2]))
11
+ if previous is not None and (np.linalg.norm(c[:2]-target[:2])>25 or abs(c[2]-target[2])>10):return None
12
+ return c.copy()
agent_d20_e20/frozen_20261102/tasks/DualArmMahjongSetup-v1/frozen/bank.npz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b9c27b3dcd2882b12be3947e6d221232011eee86b5e0f02ac92ee3599fb3cf03
3
+ size 21799399