BridgeData2_LeRobot_v3

BridgeData2 LeRobot v3 Dataset Summary `nvidia/bridge_lerobot_v3` is a LeRobotDataset v3.0 conversion of the BridgeDataset / BridgeData V2 robot manipulation dataset. BridgeData V2 is a large-scale real-world robotics dataset collected to support scalable robot learning, including imitation learning, offline reinforcement learning, and open-vocabulary multi-task policies conditioned on goal images or natural-language instructions. This repository packages Bridge trajectories in the LeRobot v3 layout with Parquet state/action data, MP4 video observations, and structured LeRobot metadata. The converted repository's `meta/info.json` reports 50,415 episodes, 1,801,090 frames, 22,199 task IDs, and a 5 Hz sampling rate. Dataset Details Dataset Description BridgeData V2 contains robot manipulation behaviors collected across varied tabletop and toy-kitchen environments. The upstream dataset was designed to study generalization across tasks, objects, environments, institutions, and conditioning modes such as language instructions and goal images. This LeRobot v3 conversion preserves the dataset for workflows that use the LeRobot data loader and Hugging Face Hub-native robotics dataset conventions. The conversion changes the storage layout: frame-level tabular data is stored in Apache Parquet shards, visual observations are stored as MP4 videos, and metadata records the schema, feature statistics, task IDs, episode boundaries, and path templates. This dataset is ready for commercial or non-commercial uses. Dataset Owner(s) NVIDIA Corporation Dataset Creation Date 2026-05-26 Version: v1.0 License/Terms of Use This dataset is released under the OpenMDW1.1 Dataset Sources Repository: https://huggingface.co/datasets/nvidia/bridge_lerobot_v3 BridgeData V2 project page: https://rail-berkeley.github.io/bridgedata/ BridgeData V2 paper: https://arxiv.org/abs/2308.12952 LeRobotDataset v3.0 documentation: https://huggingface.co/docs/lerobot/lerobot-dataset-v3 Intended Use This dataset i