Smokepord commited on
Commit
d3c3a85
·
verified ·
1 Parent(s): a621939

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +194 -0
README.md ADDED
@@ -0,0 +1,194 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ task_categories:
3
+ - robotics
4
+ tags:
5
+ - LeRobot
6
+ - piper
7
+ - pick-and-place
8
+ - imitation-learning
9
+ - real-robot
10
+ - realsense
11
+ configs:
12
+ - config_name: default
13
+ data_files: data/*/*.parquet
14
+ ---
15
+
16
+ This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
17
+
18
+
19
+ <a class="flex" href="https://huggingface.co/spaces/lerobot/visualize_dataset?path=Smokepord/piper_pick_place_xy_random_50_v1">
20
+ <img class="block dark:hidden" src="https://huggingface.co/datasets/huggingface/badges/resolve/main/visualize-this-dataset-xl.svg"/>
21
+ <img class="hidden dark:block" src="https://huggingface.co/datasets/huggingface/badges/resolve/main/visualize-this-dataset-xl-dark.svg"/>
22
+ </a>
23
+
24
+
25
+ ## Dataset Description
26
+
27
+ This dataset contains 50 successful real-robot demonstrations of an AgileX Piper arm picking up a blue 3D-printed container twist-lock part and placing it inside a black container.
28
+
29
+ The dataset contains 18,884 frames at 20 Hz, synchronized wrist and fixed-view RGB observations from two Intel RealSense D435i cameras (640 x 480, AV1), and 7-dimensional robot state/action vectors ordered as six joint positions followed by gripper position. Actions are absolute position targets in the same normalized coordinate system as the state.
30
+
31
+ The object was randomly translated within a small planar XY region. Object height and orientation, container pose, cameras, lighting, and the wider workspace arrangement remained effectively fixed. The dataset therefore supports an in-region imitation-learning baseline and should not be interpreted as covering changes in height, orientation, lighting, camera pose, container geometry, or larger workspace regions.
32
+
33
+ An ACT policy selected at checkpoint 25,000 achieved 16 successes in 20 real-robot trials (80%; approximate 95% Wilson interval: 58.4%-91.9%). Evaluation used four in-region object start positions with five trials per position; each position produced four successes and one failure. Failure stages and trial durations were not separately recorded.
34
+
35
+
36
+ - **Homepage:** [More Information Needed]
37
+ - **Paper:** [More Information Needed]
38
+ - **License:** [More Information Needed]
39
+
40
+ ## Dataset Structure
41
+
42
+ [meta/info.json](meta/info.json):
43
+ ```json
44
+ {
45
+ "codebase_version": "v3.0",
46
+ "fps": 20,
47
+ "features": {
48
+ "action": {
49
+ "dtype": "float32",
50
+ "names": [
51
+ "joint1.pos",
52
+ "joint2.pos",
53
+ "joint3.pos",
54
+ "joint4.pos",
55
+ "joint5.pos",
56
+ "joint6.pos",
57
+ "gripper.pos"
58
+ ],
59
+ "shape": [
60
+ 7
61
+ ]
62
+ },
63
+ "observation.state": {
64
+ "dtype": "float32",
65
+ "names": [
66
+ "joint1.pos",
67
+ "joint2.pos",
68
+ "joint3.pos",
69
+ "joint4.pos",
70
+ "joint5.pos",
71
+ "joint6.pos",
72
+ "gripper.pos"
73
+ ],
74
+ "shape": [
75
+ 7
76
+ ]
77
+ },
78
+ "observation.images.wrist": {
79
+ "dtype": "video",
80
+ "shape": [
81
+ 480,
82
+ 640,
83
+ 3
84
+ ],
85
+ "names": [
86
+ "height",
87
+ "width",
88
+ "channels"
89
+ ],
90
+ "info": {
91
+ "is_depth_map": false,
92
+ "video.height": 480,
93
+ "video.width": 640,
94
+ "video.codec": "av1",
95
+ "video.pix_fmt": "yuv420p",
96
+ "video.fps": 20,
97
+ "video.channels": 3,
98
+ "has_audio": false,
99
+ "video.g": 2,
100
+ "video.crf": 30,
101
+ "video.preset": 12,
102
+ "video.fast_decode": 0,
103
+ "video.video_backend": "pyav",
104
+ "video.extra_options": {}
105
+ }
106
+ },
107
+ "observation.images.fixed": {
108
+ "dtype": "video",
109
+ "shape": [
110
+ 480,
111
+ 640,
112
+ 3
113
+ ],
114
+ "names": [
115
+ "height",
116
+ "width",
117
+ "channels"
118
+ ],
119
+ "info": {
120
+ "is_depth_map": false,
121
+ "video.height": 480,
122
+ "video.width": 640,
123
+ "video.codec": "av1",
124
+ "video.pix_fmt": "yuv420p",
125
+ "video.fps": 20,
126
+ "video.channels": 3,
127
+ "has_audio": false,
128
+ "video.g": 2,
129
+ "video.crf": 30,
130
+ "video.preset": 12,
131
+ "video.fast_decode": 0,
132
+ "video.video_backend": "pyav",
133
+ "video.extra_options": {}
134
+ }
135
+ },
136
+ "timestamp": {
137
+ "dtype": "float32",
138
+ "shape": [
139
+ 1
140
+ ],
141
+ "names": null
142
+ },
143
+ "frame_index": {
144
+ "dtype": "int64",
145
+ "shape": [
146
+ 1
147
+ ],
148
+ "names": null
149
+ },
150
+ "episode_index": {
151
+ "dtype": "int64",
152
+ "shape": [
153
+ 1
154
+ ],
155
+ "names": null
156
+ },
157
+ "index": {
158
+ "dtype": "int64",
159
+ "shape": [
160
+ 1
161
+ ],
162
+ "names": null
163
+ },
164
+ "task_index": {
165
+ "dtype": "int64",
166
+ "shape": [
167
+ 1
168
+ ],
169
+ "names": null
170
+ }
171
+ },
172
+ "total_episodes": 50,
173
+ "total_frames": 18884,
174
+ "total_tasks": 1,
175
+ "chunks_size": 1000,
176
+ "data_files_size_in_mb": 100,
177
+ "video_files_size_in_mb": 200,
178
+ "data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
179
+ "video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
180
+ "robot_type": "piper_follower",
181
+ "splits": {
182
+ "train": "0:50"
183
+ }
184
+ }
185
+ ```
186
+
187
+
188
+ ## Citation
189
+
190
+ **BibTeX:**
191
+
192
+ ```bibtex
193
+ [More Information Needed]
194
+ ```