justintiensmith/VLA_Reasoning_Training_Dataset_1200_2cam

Robotics Dataset

A large-scale robot learning dataset with 1200 episodes of dual-camera (middle and wrist) RGB observations and 6-DoF joint actions, collected from a SO-100 follower robot, designed for vision-language-action (VLA) reasoning training.

Catalog metadata

Repository owner: justintiensmith
Source: Hugging Face / LeRobot
Robot: so_follower
Episodes: 1200
Frames: 612733
Cameras: 2
Frames per second: 30
License: apache-2.0

Browse the catalog to compare robots, tasks, sensors, and licenses. Sign in to open protected download and model links.