pranavsaroha commited on
Commit
5438b75
·
verified ·
1 Parent(s): 85c85a9

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +125 -0
README.md ADDED
@@ -0,0 +1,125 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ license: mit
5
+ multimodality:
6
+ - video
7
+ paperswithcode_id: null
8
+ pretty_name: so100 Teleop Dataset
9
+ tags:
10
+ - robotics
11
+ - imitation-learning
12
+ - teleoperation
13
+ - video
14
+ task_categories:
15
+ - robotics
16
+ task_ids:
17
+ - robotics-manipulation
18
+ ---
19
+
20
+ # Dataset Card for so100 Teleop Dataset
21
+
22
+ ## Dataset Description
23
+
24
+ - **Repository:** [Add repository URL]
25
+ - **Paper:** [Add paper reference if applicable]
26
+ - **Point of Contact:** [Add contact information]
27
+
28
+ ### Dataset Summary
29
+
30
+ This dataset contains teleoperation data for so100 robot manipulation tasks. The dataset includes:
31
+
32
+ - **Videos:** RGB video recordings of robot manipulation
33
+ - **Joint States:** Per-frame joint angle data
34
+ - **Actions:** Robot joint actions derived from state differences
35
+ - **Metadata:** Episode and frame indexing information
36
+
37
+ ### Supported Tasks and Leaderboards
38
+
39
+ This dataset is designed for robot imitation learning and manipulation tasks.
40
+
41
+ ### Languages
42
+
43
+ The dataset contains English language metadata.
44
+
45
+ ## Dataset Structure
46
+
47
+ ### Data Instances
48
+
49
+ Each episode contains:
50
+ - Video frames at 24.0 FPS
51
+ - Joint state data for 6 joints
52
+ - Action data for robot control
53
+ - Timestamp information
54
+ - Episode and frame indexing
55
+
56
+ ### Data Fields
57
+
58
+ - `observation.state`: Joint angle data (float32, shape: [6])
59
+ - `action`: Robot joint actions (float32, shape: [6])
60
+ - `observation.images.main`: RGB video data (video, shape: [704, 1280, 3])
61
+ - `frame_index`: Frame index within episode (int64)
62
+ - `episode_index`: Episode identifier (int64)
63
+ - `index`: Global frame index (int64)
64
+ - `task_index`: Task identifier (int64)
65
+ - `timestamp`: Time from episode start (float32)
66
+ - `next.done`: Episode termination flag (bool)
67
+
68
+ ### Data Splits
69
+
70
+ - **Train:** 348 frames across 1 episode
71
+
72
+ ## Dataset Creation
73
+
74
+ ### Source Data
75
+
76
+ #### Initial Data Collection and Normalization
77
+
78
+ The dataset was created from teleoperation recordings of so100 robot manipulation tasks.
79
+
80
+ #### Who are the source language producers?
81
+
82
+ [Add information about data collection process]
83
+
84
+ ### Annotations
85
+
86
+ #### Annotation process
87
+
88
+ [Add information about annotation process if applicable]
89
+
90
+ #### Who are the annotators?
91
+
92
+ [Add information about annotators if applicable]
93
+
94
+ ### Personal and Sensitive Information
95
+
96
+ [Add information about personal/sensitive data handling]
97
+
98
+ ## Additional Information
99
+
100
+ ### Dataset Curators
101
+
102
+ [Add curator information]
103
+
104
+ ### Licensing Information
105
+
106
+ This dataset is licensed under the MIT License.
107
+
108
+ ### Citation Information
109
+
110
+ ```bibtex
111
+ @dataset{so100_teleop_dataset,
112
+ title = {so100 Teleop Dataset},
113
+ author = {[Add author information]},
114
+ year = {2024},
115
+ url = {[Add dataset URL]}
116
+ }
117
+ ```
118
+
119
+ ### Contributions
120
+
121
+ [Add contribution information]
122
+
123
+ ### Contact
124
+
125
+ [Add contact information]