metadata
dataset_info:
features:
- name: image
dtype: image
- name: conversations
dtype: string
splits:
- name: train
num_bytes: 139475
num_examples: 1
- name: test
num_bytes: 14212404.940594058
num_examples: 100
download_size: 14113060
dataset_size: 14351879.940594058
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
task_categories:
- robotics
This dataset is used in the paper PoseLess: Depth-Free Vision-to-Joint Control via Direct Image Mapping with VLM