dex-robot commited on
Commit
1bfae7e
·
verified ·
1 Parent(s): 309e244

Upload README_en.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README_en.md +226 -0
README_en.md ADDED
@@ -0,0 +1,226 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: other
3
+ task_categories:
4
+ - robotics
5
+ tags:
6
+ - robotics
7
+ - dexterous-manipulation
8
+ - teleoperation
9
+ - motion-capture
10
+ - human-hand
11
+ - imitation-learning
12
+ - hdf5
13
+ size_categories:
14
+ - 10GB
15
+ ---
16
+
17
+ # CAP — Dexterous Manipulation Capture Dataset (Cups Ordering, Right Hand)
18
+
19
+ CAP is a **human hand motion capture + dexterous manipulation demonstration** dataset. It records an operator performing the **cups ordering (`cups_order`)** task with the **right hand**. Data is captured with a DexCap-style pipeline and ships both **compressed trajectory HDF5 files** and **raw multi-view camera frames**, ready for motion retargeting, imitation learning, and hand-pose analysis.
20
+
21
+ - Repo: `dex-robot/cap`
22
+ - Main file: `cap.zip` (10,694,426,624 bytes ≈ 9.96 GiB)
23
+ - Files: 18,940
24
+ - Capture date: 2026-07-07
25
+ - Task: cups ordering
26
+ - Hand: right
27
+
28
+ ---
29
+
30
+ ## 1. Overview
31
+
32
+ | Item | Value |
33
+ | --- | --- |
34
+ | Task | `cups_order` (ordering cups) |
35
+ | Hand | Right |
36
+ | Capture date | 2026-07-07 |
37
+ | Recordings | 5 (16:46:34 / 16:54:35 / 16:56:54 / 16:59:04 / 17:01:24) |
38
+ | Trajectory files | 5 × `.hdf5` (one per recording) |
39
+ | Camera views | 6 (left hand / right hand / head stereo / head global RGB / head global depth) |
40
+ | Image frames | 18,915 × `.png` + 5 × `.mp4` |
41
+ | Other assets | 3 × `.json`, 6 × `.txt`, 5 × `.npy`, 1 × `.raw` |
42
+ | Archive size | 10.7 GB (10,694,426,624 bytes) |
43
+
44
+ ---
45
+
46
+ ## 2. Directory Layout
47
+
48
+ ```
49
+ cap/
50
+ ├── hdf5/ # one compressed trajectory per recording
51
+ │ ├── cups_order_right_hand_20260707_164634/
52
+ │ │ └── cups_order_right_hand_20260707_164634.hdf5
53
+ │ ├── cups_order_right_hand_20260707_165435/
54
+ │ │ └── cups_order_right_hand_20260707_165435.hdf5
55
+ │ ├── cups_order_right_hand_20260707_165654/
56
+ │ │ └── cups_order_right_hand_20260707_165654.hdf5
57
+ │ ├── cups_order_right_hand_20260707_165904/
58
+ │ │ └── cups_order_right_hand_20260707_165904.hdf5
59
+ │ └── cups_order_right_hand_20260707_170124/
60
+ │ └── cups_order_right_hand_20260707_170124.hdf5
61
+ └── raw_data/ # raw multimodal stream (images + video + calibration)
62
+ └── cups_order_right_hand_20260707_164634/
63
+ └── cups_order_right_hand_20260707_164634_extracted/
64
+ └── observe/
65
+ └── vision/
66
+ ├── left_hand_rgb/ # left wrist view (video + metadata)
67
+ ├── right_hand_rgb/ # right wrist view
68
+ ├── head_zed_left/ # head ZED left
69
+ ├── head_zed_right/ # head ZED right
70
+ ├── head_global_rgb/ # head global RGB
71
+ ├── head_global_depth/ # head global depth
72
+ └── *_frames/ # extracted PNG sequences per view
73
+ ```
74
+
75
+ ---
76
+
77
+ ## 3. Recording Sessions
78
+
79
+ | Session ID | Start time | HDF5 trajectory | Raw frames |
80
+ | --- | --- | --- | --- |
81
+ | `cups_order_right_hand_20260707_164634` | 16:46:34 | ✅ | ✅ |
82
+ | `cups_order_right_hand_20260707_165435` | 16:54:35 | ✅ | — |
83
+ | `cups_order_right_hand_20260707_165654` | 16:56:54 | ✅ | — |
84
+ | `cups_order_right_hand_20260707_165904` | 16:59:04 | ✅ | — |
85
+ | `cups_order_right_hand_20260707_170124` | 17:01:24 | ✅ | — |
86
+
87
+ > Note: the full multi-view raw frames (18,915 PNG + 5 MP4) are published with the **first session** `164634`; the other four recordings ship as compressed HDF5 trajectories.
88
+
89
+ ---
90
+
91
+ ## 4. Camera Views & Frame Counts
92
+
93
+ | View directory | PNG frames | Video |
94
+ | --- | --- | --- |
95
+ | `head_global_depth_frames` | 5,567 | ✅ |
96
+ | `right_hand_rgb_frames` | 3,091 | ✅ |
97
+ | `head_zed_right_frames` | 2,631 | ✅ |
98
+ | `left_hand_rgb_frames` | 2,565 | ✅ |
99
+ | `head_zed_left_frames` | 2,548 | ✅ |
100
+ | `head_global_rgb_frames` | 2,524 | ✅ |
101
+
102
+ Frames are named `frame_%06d.png` (zero-based, contiguous) and align by index with the MP4 in the same view directory.
103
+
104
+ ---
105
+
106
+ ## 5. Download
107
+
108
+ ### 5.1 huggingface-cli (recommended, resumable)
109
+
110
+ ```bash
111
+ pip install -U "huggingface_hub[cli]"
112
+ export HF_ENDPOINT=https://hf-mirror.com # optional mirror for CN networks
113
+ huggingface-cli download dex-robot/cap cap.zip --repo-type dataset --local-dir ./cap_data
114
+ ```
115
+
116
+ ### 5.2 Python
117
+
118
+ ```python
119
+ from huggingface_hub import hf_hub_download
120
+ p = hf_hub_download("dex-robot/cap", "cap.zip", repo_type="dataset")
121
+ print(p)
122
+ ```
123
+
124
+ ### 5.3 Direct URL
125
+
126
+ ```
127
+ https://huggingface.co/datasets/dex-robot/cap/resolve/main/cap.zip
128
+ ```
129
+
130
+ ---
131
+
132
+ ## 6. Extract & Verify
133
+
134
+ ```bash
135
+ # integrity check (no output errors = OK)
136
+ unzip -t cap.zip
137
+
138
+ # extract -> cap/
139
+ unzip -q cap.zip
140
+
141
+ du -sh cap
142
+ ```
143
+
144
+ Expected:
145
+
146
+ ```
147
+ cap/
148
+ ├── hdf5/ (5 .hdf5 files)
149
+ └── raw_data/ (18,935 files)
150
+ ```
151
+
152
+ ---
153
+
154
+ ## 7. Usage
155
+
156
+ ### 7.1 Read an HDF5 trajectory
157
+
158
+ ```python
159
+ import h5py
160
+
161
+ path = "cap/hdf5/cups_order_right_hand_20260707_164634/cups_order_right_hand_20260707_164634.hdf5"
162
+ with h5py.File(path, "r") as f:
163
+ def show(name, obj):
164
+ if isinstance(obj, h5py.Dataset):
165
+ print(f"D {name:50s} shape={obj.shape} dtype={obj.dtype}")
166
+ else:
167
+ print(f"G {name}")
168
+ f.visititems(show)
169
+ ```
170
+
171
+ ### 7.2 Read image frames
172
+
173
+ ```python
174
+ import cv2, glob
175
+
176
+ frames = sorted(glob.glob(
177
+ "cap/raw_data/cups_order_right_hand_20260707_164634/"
178
+ "cups_order_right_hand_20260707_164634_extracted/observe/vision/"
179
+ "head_global_rgb_frames/frame_*.png"))
180
+ print("frames:", len(frames))
181
+ img = cv2.imread(frames[0])
182
+ print(img.shape) # (H, W, 3)
183
+ ```
184
+
185
+ ---
186
+
187
+ ## 8. Recommended Environment
188
+
189
+ ```bash
190
+ pip install h5py opencv-python numpy
191
+ ```
192
+
193
+ - Python ≥ 3.8
194
+ - ~15 GB free disk after extraction (raw PNG sequences are large)
195
+ - `head_global_depth` may be 16-bit PNG or accompanied by `.npy` / `.raw` raw depth
196
+
197
+ ---
198
+
199
+ ## 9. License & Citation
200
+
201
+ - License: see the repo `LICENSE`; contact the publisher if unspecified.
202
+
203
+ ```bibtex
204
+ @misc{cap_dataset_2026,
205
+ title = {CAP: Cups Ordering Dexterous Manipulation Capture Dataset (Right Hand)},
206
+ author = {dex-robot},
207
+ year = {2026},
208
+ howpublished = {\url{https://huggingface.co/datasets/dex-robot/cap}}
209
+ }
210
+ ```
211
+
212
+ ---
213
+
214
+ ## 10. FAQ
215
+
216
+ **Q: Download interrupted?**
217
+ `huggingface-cli download` resumes automatically — just re-run the same command. Add `--max-workers 1` for a more stable (slower) transfer.
218
+
219
+ **Q: `unzip` says "End-of-central-directory signature not found"?**
220
+ The download is incomplete / truncated. Check that the file size is exactly **10,694,426,624 bytes**; if not, re-download.
221
+
222
+ **Q: I only need trajectories, not the 10 GB of images?**
223
+ Currently only the full bundle is published. Open a Discussion if you need a split release.
224
+
225
+ **Q: How do I align depth with RGB?**
226
+ `*_frames` and the MP4 of the same view share frame indices (`frame_%06d`). Cross-view alignment: use the calibration JSON under `raw_data/.../observe/vision/`.