Jingyi-Z commited on
Commit
3aea45c
·
verified ·
1 Parent(s): c86f7e2

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +198 -182
README.md CHANGED
@@ -1,182 +1,198 @@
1
- ---
2
- license: apache-2.0
3
- task_categories:
4
- - robotics
5
- tags:
6
- - LeRobot
7
- configs:
8
- - config_name: default
9
- data_files: data/*/*.parquet
10
- ---
11
-
12
- This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
13
-
14
-
15
- <a class="flex" href="https://huggingface.co/spaces/lerobot/visualize_dataset?path=Jingyi-Z/sotac">
16
- <img class="block dark:hidden" src="https://huggingface.co/datasets/huggingface/badges/resolve/main/visualize-this-dataset-xl.svg"/>
17
- <img class="hidden dark:block" src="https://huggingface.co/datasets/huggingface/badges/resolve/main/visualize-this-dataset-xl-dark.svg"/>
18
- </a>
19
-
20
-
21
- ## Dataset Description
22
-
23
-
24
-
25
- - **Homepage:** [More Information Needed]
26
- - **Paper:** [More Information Needed]
27
- - **License:** apache-2.0
28
-
29
- ## Dataset Structure
30
-
31
- [meta/info.json](meta/info.json):
32
- ```json
33
- {
34
- "codebase_version": "v3.0",
35
- "fps": 30,
36
- "features": {
37
- "action": {
38
- "dtype": "float32",
39
- "names": [
40
- "shoulder_pan.pos",
41
- "shoulder_lift.pos",
42
- "elbow_flex.pos",
43
- "wrist_flex.pos",
44
- "wrist_roll.pos",
45
- "gripper.pos"
46
- ],
47
- "shape": [
48
- 6
49
- ]
50
- },
51
- "observation.state": {
52
- "dtype": "float32",
53
- "names": [
54
- "shoulder_pan.pos",
55
- "shoulder_lift.pos",
56
- "elbow_flex.pos",
57
- "wrist_flex.pos",
58
- "wrist_roll.pos",
59
- "gripper.pos"
60
- ],
61
- "shape": [
62
- 6
63
- ]
64
- },
65
- "observation.images.wrist": {
66
- "dtype": "video",
67
- "shape": [
68
- 480,
69
- 640,
70
- 3
71
- ],
72
- "names": [
73
- "height",
74
- "width",
75
- "channels"
76
- ],
77
- "info": {
78
- "video.height": 480,
79
- "video.width": 640,
80
- "video.codec": "av1",
81
- "video.pix_fmt": "yuv420p",
82
- "video.is_depth_map": false,
83
- "video.fps": 30,
84
- "video.channels": 3,
85
- "has_audio": false
86
- }
87
- },
88
- "observation.images.top": {
89
- "dtype": "video",
90
- "shape": [
91
- 480,
92
- 640,
93
- 3
94
- ],
95
- "names": [
96
- "height",
97
- "width",
98
- "channels"
99
- ],
100
- "info": {
101
- "video.height": 480,
102
- "video.width": 640,
103
- "video.codec": "av1",
104
- "video.pix_fmt": "yuv420p",
105
- "video.is_depth_map": false,
106
- "video.fps": 30,
107
- "video.channels": 3,
108
- "has_audio": false
109
- }
110
- },
111
- "observation.sensors.paxini_fingertip": {
112
- "dtype": "float32",
113
- "shape": [
114
- 2,
115
- 52,
116
- 3
117
- ],
118
- "names": [
119
- "dim_0",
120
- "dim_1",
121
- "dim_2"
122
- ]
123
- },
124
- "timestamp": {
125
- "dtype": "float32",
126
- "shape": [
127
- 1
128
- ],
129
- "names": null
130
- },
131
- "frame_index": {
132
- "dtype": "int64",
133
- "shape": [
134
- 1
135
- ],
136
- "names": null
137
- },
138
- "episode_index": {
139
- "dtype": "int64",
140
- "shape": [
141
- 1
142
- ],
143
- "names": null
144
- },
145
- "index": {
146
- "dtype": "int64",
147
- "shape": [
148
- 1
149
- ],
150
- "names": null
151
- },
152
- "task_index": {
153
- "dtype": "int64",
154
- "shape": [
155
- 1
156
- ],
157
- "names": null
158
- }
159
- },
160
- "total_episodes": 77,
161
- "total_frames": 42684,
162
- "total_tasks": 3,
163
- "chunks_size": 1000,
164
- "data_files_size_in_mb": 100,
165
- "video_files_size_in_mb": 200,
166
- "data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
167
- "video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
168
- "robot_type": "so_sensor_follower",
169
- "splits": {
170
- "train": "0:77"
171
- }
172
- }
173
- ```
174
-
175
-
176
- ## Citation
177
-
178
- **BibTeX:**
179
-
180
- ```bibtex
181
- [More Information Needed]
182
- ```
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - robotics
5
+ tags:
6
+ - LeRobot
7
+ - tactile
8
+ - so101
9
+ - paxini
10
+ configs:
11
+ - config_name: default
12
+ data_files: data/*/*.parquet
13
+ language:
14
+ - en
15
+ pretty_name: SoTac
16
+ size_categories:
17
+ - 10K<n<100K
18
+ ---
19
+
20
+ # SoTac — SO-101 Tactile Manipulation Corpus
21
+
22
+ SoTac is a growing corpus of visuo-tactile teleoperation demonstrations recorded on an
23
+ SO-101 arm with dual 3-axis tactile fingertip sensors. Every 30 Hz observation carries a
24
+ full per-taxel force map for both fingertips, and every episode ships with the raw
25
+ ~91 Hz tactile stream as a sidecar — no downsampling at record time.
26
+
27
+ Created using [LeRobot](https://github.com/huggingface/lerobot) (via the
28
+ [lerobotac](https://github.com/Jingyi-Z/lerobotac) sensor fork).
29
+
30
+ <a class="flex" href="https://jingyi-z-lerobotac-dataset-visualizer.hf.space/Jingyi-Z/sotac/0">
31
+ <img class="block dark:hidden" src="https://huggingface.co/datasets/huggingface/badges/resolve/main/visualize-this-dataset-xl.svg"/>
32
+ <img class="hidden dark:block" src="https://huggingface.co/datasets/huggingface/badges/resolve/main/visualize-this-dataset-xl-dark.svg"/>
33
+ </a>
34
+
35
+ Use the button above — it opens the [LeRobotAC tactile visualizer](https://huggingface.co/spaces/Jingyi-Z/lerobotac-dataset-visualizer)
36
+ with per-taxel 3D force arrows, contact timelines, and the raw 91 Hz stream panels.
37
+ The standard LeRobot viewer does not render the tactile data.
38
+
39
+ ## Contents
40
+
41
+ SoTac is actively growing — new tasks, objects, and sensors will be added over time.
42
+
43
+ Current release (2026-08-18): 62 episodes / 35,412 frames (~20 min) across
44
+ 3 pick-and-place tasks:
45
+
46
+ | task | episodes | episode # |
47
+ |---|---|---|
48
+ | Pick up the red foam ball and place it into the container | 25 | 0–9, 47–61 |
49
+ | Pick up the orange rubber ball and place it into the container | 17 | 10–26 |
50
+ | Pick up the red cup and place it on top of the container | 20 | 27–46 |
51
+
52
+ Grip forces span gentle (~4–10 N peak) to firm (~50 N peak) handling across sessions.
53
+
54
+ **Data collection**: episodes 0–9 & 27–31 by Jingming Zhang, episodes 10–26 by
55
+ Yuzhou Wang, episodes 32–61 by Jingyi Zou.
56
+
57
+ ## Hardware
58
+
59
+ - **Robot**: SO-101 follower arm, teleoperated with an SO-101 leader arm
60
+ - **Cameras**: wrist + top, 640×480 @ 30 fps (MJPG)
61
+ - **Tactile**: 2× Paxini PX-6AX GEN3 (DP-S2015-Elite) fingertip pads on a shared
62
+ High-Speed Communication Board — 52 taxels per pad, 3-axis force per taxel,
63
+ 0.1 N/LSB, ~91 Hz effective rate for both pads together. Firmware zeroing at
64
+ connect. Finger `[0]` = gripper-side pad (module 10), finger `[1]` =
65
+ wrist-roll-side pad (module 18).
66
+
67
+ ## Data format
68
+
69
+ Standard [LeRobot](https://github.com/huggingface/lerobot) v3.0 dataset, plus one
70
+ extra feature and one extra directory:
71
+
72
+ **`observation.sensors.paxini_fingertip`** — float32, shape `(2, 52, 3)`:
73
+ the latest raw tactile sample per finger at each 30 Hz observation
74
+ (fingers × taxels × [fx, fy, fz], in units of 0.1 N). Per-taxel (x, y, z) mm
75
+ coordinates for the GEN3 layout are available in
76
+ [paxini-sdk](https://github.com/Jingyi-Z/paxini-sdk).
77
+
78
+ **`sensors/paxini_fingertip/episode_XXXXXX/`** — raw ~91 Hz stream sidecars:
79
+
80
+ - `sensor_1.csv`, `sensor_2.csv` (fingers 0 and 1): 163 columns —
81
+ `timestamp_ns, frame_status, time_calibration_offset_ns, calibrated_timestamp_ns,
82
+ fx, fy, fz, p_00_fx … p_51_fz` (resultant force followed by 52 per-taxel triplets)
83
+ - `alignment.json`: episode-start epoch timestamp (`time.time_ns`) anchoring the raw
84
+ stream to frame 0 of the 30 Hz main table
85
+
86
+ ## Collection protocol
87
+
88
+ Episodes start with the gripper open and the pads contact-free (calibration reference).
89
+ The operator and leader arm are kept out of both camera views. Object placement is
90
+ varied between episodes.
91
+
92
+ ## Tooling
93
+
94
+ - Recording: [lerobotac](https://github.com/Jingyi-Z/lerobotac) (`hall-sensor` branch) —
95
+ LeRobot fork with a sensor framework; the `paxini` sensor's `combined` output mode +
96
+ `record_raw_csv` produced this dataset
97
+ - Sensor SDK: [paxini-sdk](https://github.com/Jingyi-Z/paxini-sdk)
98
+ - Visualization / annotation: [LeRobotAC Dataset Visualizer](https://huggingface.co/spaces/Jingyi-Z/lerobotac-dataset-visualizer)
99
+
100
+ ## Dataset Structure
101
+
102
+ [meta/info.json](meta/info.json):
103
+
104
+ ```json
105
+ {
106
+ "codebase_version": "v3.0",
107
+ "fps": 30,
108
+ "features": {
109
+ "action": {
110
+ "dtype": "float32",
111
+ "names": [
112
+ "shoulder_pan.pos",
113
+ "shoulder_lift.pos",
114
+ "elbow_flex.pos",
115
+ "wrist_flex.pos",
116
+ "wrist_roll.pos",
117
+ "gripper.pos"
118
+ ],
119
+ "shape": [6]
120
+ },
121
+ "observation.state": {
122
+ "dtype": "float32",
123
+ "names": [
124
+ "shoulder_pan.pos",
125
+ "shoulder_lift.pos",
126
+ "elbow_flex.pos",
127
+ "wrist_flex.pos",
128
+ "wrist_roll.pos",
129
+ "gripper.pos"
130
+ ],
131
+ "shape": [6]
132
+ },
133
+ "observation.images.wrist": {
134
+ "dtype": "video",
135
+ "shape": [480, 640, 3],
136
+ "names": ["height", "width", "channels"],
137
+ "info": {
138
+ "video.height": 480,
139
+ "video.width": 640,
140
+ "video.codec": "av1",
141
+ "video.pix_fmt": "yuv420p",
142
+ "video.is_depth_map": false,
143
+ "video.fps": 30,
144
+ "video.channels": 3,
145
+ "has_audio": false
146
+ }
147
+ },
148
+ "observation.images.top": {
149
+ "dtype": "video",
150
+ "shape": [480, 640, 3],
151
+ "names": ["height", "width", "channels"],
152
+ "info": {
153
+ "video.height": 480,
154
+ "video.width": 640,
155
+ "video.codec": "av1",
156
+ "video.pix_fmt": "yuv420p",
157
+ "video.is_depth_map": false,
158
+ "video.fps": 30,
159
+ "video.channels": 3,
160
+ "has_audio": false
161
+ }
162
+ },
163
+ "observation.sensors.paxini_fingertip": {
164
+ "dtype": "float32",
165
+ "shape": [2, 52, 3],
166
+ "names": ["finger", "taxel", "axis"]
167
+ },
168
+ "timestamp": {"dtype": "float32", "shape": [1], "names": null},
169
+ "frame_index": {"dtype": "int64", "shape": [1], "names": null},
170
+ "episode_index": {"dtype": "int64", "shape": [1], "names": null},
171
+ "index": {"dtype": "int64", "shape": [1], "names": null},
172
+ "task_index": {"dtype": "int64", "shape": [1], "names": null}
173
+ },
174
+ "total_episodes": 62,
175
+ "total_frames": 35412,
176
+ "total_tasks": 3,
177
+ "chunks_size": 1000,
178
+ "data_files_size_in_mb": 100,
179
+ "video_files_size_in_mb": 200,
180
+ "data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
181
+ "video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
182
+ "robot_type": "so_sensor_follower",
183
+ "splits": {
184
+ "train": "0:62"
185
+ }
186
+ }
187
+ ```
188
+
189
+ ## Citation
190
+
191
+ ```bibtex
192
+ @misc{zou2026sotac,
193
+ author = {Zou, Jingyi},
194
+ title = {SoTac: An SO-101 Tactile Manipulation Corpus},
195
+ year = {2026},
196
+ howpublished = {\url{https://huggingface.co/datasets/Jingyi-Z/sotac}}
197
+ }
198
+ ```