A large-scale robotics dataset for vision-language-action learning, featuring 791 datasets across 46 robot types, enabling cross-embodiment pretraining for generalist robot policies.
This is a crowdsourced, open-source dataset compiled from 235 community contributors worldwide. Building upon the pretraining datasets used for SmolVLA, Community… See the full description on the dataset page:
https://huggingface.co/datasets/lerobot/community_dataset_v3.
