Hugging Face's logo

Datasets:
ALRpraktikum
/
Teachingtopack

Tasks:
Summarization
Size:
100M<n<1B
Tags:
code
License:
Dataset card Files Community
Teachingtopack / README.md
R4who's picture
R4who
Update README.md
406ef68 verified
preview code
|
raw
history blame contribute delete
1.16 kB
metadata
license: mit
task_categories:
  - summarization
tags:
  - code
pretty_name: aloha_robots_hfd5_compressed
size_categories:
  - 100M<n<1B

Dataset Card for Teaching to pack

this dataset contains different demonstrations recorded from aloha robots. different tasks may have different amounts demonstrations.

Dataset Details

Dataset Description

the datasets are hdf5 files and its format be like : """ For each timestep: observations - images - cam_high (480, 640, 3) 'uint8' - cam_low (480, 640, 3) 'uint8' - cam_left_wrist (480, 640, 3) 'uint8' - cam_right_wrist (480, 640, 3) 'uint8' - qpos (14,) 'float64' - qvel (14,) 'float64'

action                  (14,)         'float64'
"""

Dataset Sources [optional]

Uses

Files are all compressed, the images of the observation need to be decoded before using """decompressed_image = cv2.imdecode(image_dict[cam_name], 1)"""