Hugging Face's logo

Datasets:
nvidia
/
PhysicalAI-Robotics-GR00T-Teleop-G1

Modalities:
Tabular
Video
Formats:
parquet
Size:
100K - 1M
Libraries:
Datasets
Dask
License:
Dataset card Data Studio Files
xet
Community
2
youliangt commited on
Commit
c0bdd6c
·
verified ·
1 Parent(s): 97f537a

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +35 -4
README.md CHANGED
@@ -1,7 +1,38 @@
 
1
  ## Unitree G1 Fruits Pick and Place 1K Dataset
2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
 
4
- ![image/png](https://cdn-uploads.huggingface.co/production/uploads/67b8da81d01134f89899b4a7/y8-gdvfIRhKtxbcQCYYRv.png)
5
- ![image/png](https://cdn-uploads.huggingface.co/production/uploads/67b8da81d01134f89899b4a7/zYqxhAGHV26raIAh9p1BS.png)
6
- ![image/png](https://cdn-uploads.huggingface.co/production/uploads/67b8da81d01134f89899b4a7/a2ePP_DycQl7cjoV86QR6.png)
7
- ![image/png](https://cdn-uploads.huggingface.co/production/uploads/67b8da81d01134f89899b4a7/2260l3TdVqCM3R9FOCAMo.png)
 
1
+
2
  ## Unitree G1 Fruits Pick and Place 1K Dataset
3
 
4
+ This dataset is a collection of total 1K episodes of fruits picking and placing tasks on a Unitree G1 robot.
5
+
6
+ The robot is equipped with the default realsense camera, and a pair of Unitree G1 Tri-fingers hand.
7
+
8
+
9
+ Modality:
10
+ - Observation: 43 dim of vectorized state (joint positions of full body + hands)
11
+ - Action: 43 dim of vectorized action (joint positions of full body + hands)
12
+ - Video: RGB video, 640x480 resolution, 20fps
13
+ - Language Instruction:
14
+ - *"Pick the apple from the table and place it into the basket."*
15
+ - *"Pick the pear from the table and place it into the basket."*
16
+ - *"Pick the grapes from the table and place it into the basket."*
17
+ - *"Pick the starfruit from the table and place it into the basket."*
18
+
19
+
20
+ | Dataset Folder | Preview |
21
+ |----------------|---------|
22
+ | g1-pick-apple | ![image/png](https://cdn-uploads.huggingface.co/production/uploads/67b8da81d01134f89899b4a7/y8-gdvfIRhKtxbcQCYYRv.png) |
23
+ | g1-pick-pear | ![image/png](https://cdn-uploads.huggingface.co/production/uploads/67b8da81d01134f89899b4a7/zYqxhAGHV26raIAh9p1BS.png) |
24
+ | g1-pick-grapes | ![image/png](https://cdn-uploads.huggingface.co/production/uploads/67b8da81d01134f89899b4a7/a2ePP_DycQl7cjoV86QR6.png) |
25
+ | g1-pick-starfruit | ![image/png](https://cdn-uploads.huggingface.co/production/uploads/67b8da81d01134f89899b4a7/2260l3TdVqCM3R9FOCAMo.png) |
26
+
27
+
28
+ ## Download the dataset
29
+
30
+ ```bash
31
+ huggingface-cli download \
32
+ --repo-type dataset nvidia/PhysicalAI-Robotics-GR00T-Teleop-G1 \
33
+ --local-dir ./datasets/
34
+ ```
35
+
36
+ ## Finetuning with GR00T-N1.5
37
 
38
+ Refer to the [Github repo](https://github.com/NVIDIA/Isaac-GR00T/blob/main/getting_started/3_0_new_embodiment_finetuning.md) for the finetuning script.