Real-world data for AgiBot World 2026 robots
Corpus of real scenes with images and video for training and evaluating embodied AI and robotics.
- Robotics
- English
Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,788 in the catalog so far, with new ones added every hour.
Answered by Gemini 4 · the chat is kept only in this tab
Corpus of real scenes with images and video for training and evaluating embodied AI and robotics.
Large multimodal corpus of human hand manipulation videos with motion annotations and descriptions for robot learning.
Large teleoperation video dataset of a bimanual robot with detailed manual step labels for training robotics models.
Large-scale robot teleoperation recordings across 56 tasks using head and wrist cameras.
Large video dataset of robot interactions with the real environment for reinforcement learning.
Multimodal web interaction recordings formatted as dialogs for training web agents.
Large egocentric dataset of hand action recordings in Chinese and English.
A CVPR 2025 dataset for video object segmentation under phase changes, transitions, and scene changes.
Over 10,000 hours of real-world recordings of craftsmanship, cooking, and construction work with IMU sensor data.
139.4 hours of first-person teleoperation video recordings for training robots in object manipulation.
Robotwin dataset in LeRobot format with pre-extracted video latents for fine-tuning robotics models.
Over 215 hours of recordings from 41 cameras and tactile grasps interacting with 198 deformable objects.
Robot manipulation video corpus: 13k hours, 5M clips, and everyday scenarios for embodied AI training.
4,000+ hours of audio and video of in-person interactions by 4,000+ participants for AI research.
A large-scale dataset of real bimanual robot video and telemetry with text descriptions for 90 everyday tasks.
Unedited daily and working first-person video recordings in English for training video models.
87k video episodes and telemetry of 599 robot tasks for training robotics models.
Multimodal dataset with video, 3D hand and object poses in real-world conditions for CVPR 2026.
Over 53,000 episodes of WidowX robot operation for training control systems.
Largest first-person manual labor video dataset: 100,405 hours and 2 million clips.
Large multimodal dataset of 1M to 10M examples for training robots in safe human interaction.
Synchronized video, audio, and 3D data from real home environments for embodied AI training.
Over 8,100 English and Chinese prompts together with generated videos and preview frames.
Multi-view video recordings of toy vehicle assembly with action and error annotations.