EXCALORIO commited on
Commit
46bb468
·
verified ·
1 Parent(s): fc257d3

ficha: que se extrae, con que precision, y los videos

Browse files
Files changed (1) hide show
  1. README.md +63 -0
README.md CHANGED
@@ -44,10 +44,39 @@ us. If you work on fencing video, start there: **[fencingdatabase.com](https://f
44
  Every video below is the **same clip**, one layer at a time, at the clip's own frame rate. The label
45
  on each frame says what you are looking at and what it measured.
46
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
47
  ### 1 - POSE: 308 keypoints per fencer
48
 
49
  Two fencers, 308 points each, every frame. Green is side 0 (image left), orange is side 1. A red box is a person the detector found and we **did not** select. The insets at the bottom are the weapon hand at native resolution, because each hand is 21 of those 308 points and at full frame they are a blur.
50
 
 
 
 
 
51
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/pose_BabyishIdioticCockerspaniel.mp4" controls loop muted playsinline width="100%"></video>
52
 
53
  [pose_BabyishIdioticCockerspaniel.mp4 - MP4, 3.1 MB](media/pose_BabyishIdioticCockerspaniel.mp4)
@@ -79,6 +108,10 @@ More stills, same layer: [BabyishIdioticCockerspaniel_c0302](media/pose_BabyishI
79
 
80
  The cleaned mask, blended **only inside itself**. Outside it the frame is just darkened, which is deliberate: that is how you can check the mask landed on the right person instead of on the referee walking past.
81
 
 
 
 
 
82
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/seg_BabyishIdioticCockerspaniel.mp4" controls loop muted playsinline width="100%"></video>
83
 
84
  [seg_BabyishIdioticCockerspaniel.mp4 - MP4, 2.1 MB](media/seg_BabyishIdioticCockerspaniel.mp4)
@@ -106,6 +139,10 @@ More stills, same layer: [BabyishIdioticCockerspaniel_c0302](media/seg_BabyishId
106
 
107
  One axis per colour channel. This is the dense layer that **does** hold up for local geometry.
108
 
 
 
 
 
109
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/normal_BabyishIdioticCockerspaniel.mp4" controls loop muted playsinline width="100%"></video>
110
 
111
  [normal_BabyishIdioticCockerspaniel.mp4 - MP4, 2.1 MB](media/normal_BabyishIdioticCockerspaniel.mp4)
@@ -131,6 +168,10 @@ More stills, same layer: [BabyishIdioticCockerspaniel_c0302](media/normal_Babyis
131
 
132
  The soft silhouette. At very low alpha the **edge of the blade** appears, and that is what the foil detector works on.
133
 
 
 
 
 
134
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/matting_BabyishIdioticCockerspaniel.mp4" controls loop muted playsinline width="100%"></video>
135
 
136
  [matting_BabyishIdioticCockerspaniel.mp4 - MP4, 1.9 MB](media/matting_BabyishIdioticCockerspaniel.mp4)
@@ -156,6 +197,10 @@ More stills, same layer: [BabyishIdioticCockerspaniel_c0302](media/matting_Babyi
156
 
157
  Shown because it was asked for, not because it is used. It is kept as a channel and the pipeline ignores it.
158
 
 
 
 
 
159
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/pointmap_BabyishIdioticCockerspaniel.mp4" controls loop muted playsinline width="100%"></video>
160
 
161
  [pointmap_BabyishIdioticCockerspaniel.mp4 - MP4, 2.2 MB](media/pointmap_BabyishIdioticCockerspaniel.mp4)
@@ -186,6 +231,10 @@ The cone is what turns *where the arm points* into *who can be hit*. It is **not
186
 
187
  The yellow cross is where **the two cones cross each other**. That is the test - cone against cone, not cone against torso.
188
 
 
 
 
 
189
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/foil_BabyishIdioticCockerspaniel.mp4" controls loop muted playsinline width="100%"></video>
190
 
191
  [foil_BabyishIdioticCockerspaniel.mp4 - MP4, 2.7 MB](media/foil_BabyishIdioticCockerspaniel.mp4)
@@ -221,6 +270,20 @@ This was rebuilt from scratch, because the first version had a defect worth stat
221
 
222
  Left: the 308-point skeleton with each fencer's position in metres. Right: the trace - position along the piste against time, which is the racing-line view. Bottom: the piste from above, to scale, with the regulation lines and the events appearing as they happen. The counter is there so the count can be judged frame by frame rather than taken on trust.
223
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
224
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/trace_ForcefulAntiqueAlbacoretuna_public.mp4" controls loop muted playsinline width="100%"></video>
225
 
226
  [trace_ForcefulAntiqueAlbacoretuna_public.mp4 - MP4, 7.0 MB](media/trace_ForcefulAntiqueAlbacoretuna_public.mp4)
 
44
  Every video below is the **same clip**, one layer at a time, at the clip's own frame rate. The label
45
  on each frame says what you are looking at and what it measured.
46
 
47
+ ### How a pose gets found: four rounds, and what each one recovers
48
+
49
+ A pose is not simply *detected or not*. There are four rounds, and **each one is counted separately** so you can see what would be missing without it.
50
+
51
+ **The pose is Sapiens 2 in every round.** What changes between rounds is **where the box comes from**: YOLO gives only the box, and the 308 points are always measured by the same pose head. Round 2 exists because that head is **top-down** - it needs no detector to confirm the person, only a box - so it can try exactly where the detector already failed twice. Round 3 is the only one with no measurement left in it.
52
+
53
+ ![cascada_deteccion](media/cascada_deteccion.png)
54
+
55
+ | round | where the **box** comes from (the pose is always Sapiens 2) | recovers | of 21,950 fencer-frames |
56
+ |---|---|---:|---:|
57
+ | **0** | YOLO on the whole frame | 21,274 | 96.92 % |
58
+ | **1** | **YOLO again, on a crop** - the redetection | 252 | 1.15 % |
59
+ | **2** | **interpolated box, no detector at all** | 50 | 0.23 % |
60
+ | **3** | no box and no pose: **the pose itself is interpolated** | 600 | 2.73 % |
61
+
62
+ Read it as a cascade: **without round 1 the pose would be missing in 3.08 %** of fencer-frames. Round 1 - the redetection - recovers **252 of them**, leaving **2.96 %** with no measured pose. Round 2 recovers **50 more**, and what is left, **2.73 %**, is interpolated and **carries a flag saying so**.
63
+
64
+ Round 2 is only attempted where the gap is **24 frames or shorter** (50 of 295 cases). Beyond that it is not a tuning problem: the quality gate passes **0.6 %** of the time with a fencer in the box and **12.7 %** on an empty patch of background, so it does not discriminate at all - the fencer is simply not in those frames.
65
+
66
+ **A redetection you can look at.** Left: the full frame, with the box where the detector found nobody. Right: the same box as a crop, where the pose came out.
67
+
68
+ ![redeteccion_01_FirsthandSoreGnat_der_c0055](media/redeteccion_01_FirsthandSoreGnat_der_c0055.png)
69
+ ![redeteccion_02_UnevenChiefHippopotamu_der_c0023](media/redeteccion_02_UnevenChiefHippopotamu_der_c0023.png)
70
+
71
+
72
  ### 1 - POSE: 308 keypoints per fencer
73
 
74
  Two fencers, 308 points each, every frame. Green is side 0 (image left), orange is side 1. A red box is a person the detector found and we **did not** select. The insets at the bottom are the weapon hand at native resolution, because each hand is 21 of those 308 points and at full frame they are a blur.
75
 
76
+ ![capa_pose](media/capa_pose.gif)
77
+
78
+ *8 s of the layer, 720 px. The MP4 below is the whole clip at full resolution - the GIF is here because GitHub does not play the video.*
79
+
80
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/pose_BabyishIdioticCockerspaniel.mp4" controls loop muted playsinline width="100%"></video>
81
 
82
  [pose_BabyishIdioticCockerspaniel.mp4 - MP4, 3.1 MB](media/pose_BabyishIdioticCockerspaniel.mp4)
 
108
 
109
  The cleaned mask, blended **only inside itself**. Outside it the frame is just darkened, which is deliberate: that is how you can check the mask landed on the right person instead of on the referee walking past.
110
 
111
+ ![capa_seg](media/capa_seg.gif)
112
+
113
+ *8 s of the layer, 720 px. The MP4 below is the whole clip at full resolution - the GIF is here because GitHub does not play the video.*
114
+
115
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/seg_BabyishIdioticCockerspaniel.mp4" controls loop muted playsinline width="100%"></video>
116
 
117
  [seg_BabyishIdioticCockerspaniel.mp4 - MP4, 2.1 MB](media/seg_BabyishIdioticCockerspaniel.mp4)
 
139
 
140
  One axis per colour channel. This is the dense layer that **does** hold up for local geometry.
141
 
142
+ ![capa_normal](media/capa_normal.gif)
143
+
144
+ *8 s of the layer, 720 px. The MP4 below is the whole clip at full resolution - the GIF is here because GitHub does not play the video.*
145
+
146
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/normal_BabyishIdioticCockerspaniel.mp4" controls loop muted playsinline width="100%"></video>
147
 
148
  [normal_BabyishIdioticCockerspaniel.mp4 - MP4, 2.1 MB](media/normal_BabyishIdioticCockerspaniel.mp4)
 
168
 
169
  The soft silhouette. At very low alpha the **edge of the blade** appears, and that is what the foil detector works on.
170
 
171
+ ![capa_matting](media/capa_matting.gif)
172
+
173
+ *8 s of the layer, 720 px. The MP4 below is the whole clip at full resolution - the GIF is here because GitHub does not play the video.*
174
+
175
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/matting_BabyishIdioticCockerspaniel.mp4" controls loop muted playsinline width="100%"></video>
176
 
177
  [matting_BabyishIdioticCockerspaniel.mp4 - MP4, 1.9 MB](media/matting_BabyishIdioticCockerspaniel.mp4)
 
197
 
198
  Shown because it was asked for, not because it is used. It is kept as a channel and the pipeline ignores it.
199
 
200
+ ![capa_pointmap](media/capa_pointmap.gif)
201
+
202
+ *8 s of the layer, 720 px. The MP4 below is the whole clip at full resolution - the GIF is here because GitHub does not play the video.*
203
+
204
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/pointmap_BabyishIdioticCockerspaniel.mp4" controls loop muted playsinline width="100%"></video>
205
 
206
  [pointmap_BabyishIdioticCockerspaniel.mp4 - MP4, 2.2 MB](media/pointmap_BabyishIdioticCockerspaniel.mp4)
 
231
 
232
  The yellow cross is where **the two cones cross each other**. That is the test - cone against cone, not cone against torso.
233
 
234
+ ![capa_foil](media/capa_foil.gif)
235
+
236
+ *8 s of the cone. The MP4 below is the whole clip at full resolution - the GIF is here because GitHub does not play the video.*
237
+
238
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/foil_BabyishIdioticCockerspaniel.mp4" controls loop muted playsinline width="100%"></video>
239
 
240
  [foil_BabyishIdioticCockerspaniel.mp4 - MP4, 2.7 MB](media/foil_BabyishIdioticCockerspaniel.mp4)
 
270
 
271
  Left: the 308-point skeleton with each fencer's position in metres. Right: the trace - position along the piste against time, which is the racing-line view. Bottom: the piste from above, to scale, with the regulation lines and the events appearing as they happen. The counter is there so the count can be judged frame by frame rather than taken on trust.
272
 
273
+ ![pista_completa](media/pista_completa.gif)
274
+
275
+ *12 s of the whole four-zone frame: skeletons and metres top-left, the label with the event underneath, the trace on the right, the piste from above along the bottom.*
276
+
277
+ The same thing zoomed into the two panels that carry the measurement, because at full frame their axis labels are too small to read:
278
+
279
+ ![pista_trazado](media/pista_trazado.gif)
280
+
281
+ ***The trace.** Position along the piste on the horizontal axis, **time** downwards. Green is the left fencer, orange the right. Both coordinates in this panel are measured: x by homography, time exactly.*
282
+
283
+ ![pista_mapa](media/pista_mapa.gif)
284
+
285
+ ***The piste from above, to scale**, with the regulation lines. Each fencer is a thick dot, the trail behind is where they have been, and the events - march, retreat, lunge, halt - appear labelled as they happen.*
286
+
287
  <video src="https://huggingface.co/datasets/EXCALORIO/fencingly-datos/resolve/main/media/trace_ForcefulAntiqueAlbacoretuna_public.mp4" controls loop muted playsinline width="100%"></video>
288
 
289
  [trace_ForcefulAntiqueAlbacoretuna_public.mp4 - MP4, 7.0 MB](media/trace_ForcefulAntiqueAlbacoretuna_public.mp4)