{"name":"Open Robot Training Datasets: Size, Collection Method and Licence (live data, 2 Jan 2023 to 18 Oct 2026)","question":"Which public datasets teach robots, embodied AI, self-driving systems and drones, who released them, how many hours and episodes do they hold and how was the data collected?","source":"Each dataset's paper, project page or dataset card, and reporting where the release gives no date or worker count, each linked with a verbatim quote kept in the dataset.","publisher":"RobotTraining.jobs","url":"https://robottraining.jobs/research/open-robot-training-datasets","license":"https://creativecommons.org/licenses/by/4.0/","citation":"RobotTraining.jobs (2026). Open Robot Training Datasets: Size, Collection Method and Licence (live data, 2 Jan 2023 to 18 Oct 2026). https://robottraining.jobs/research/open-robot-training-datasets (accessed 18 Oct 2026)","frozen_at":null,"report":"open-robot-training-datasets","method":1,"edition":null,"window":{"from":"2023-01-02","to":"2026-10-18","days":1386},"computedAt":"2026-10-18T20:07:26.197Z","summary":"32 open datasets of robot, first-person video, driving and drone training data were released from 2023 on; each is listed with who released it, its stated size, how its data was collected and its licence, linked to its source.","stats":[{"id":"datasets","value":"32","label":"open training datasets since 2023","figure":"datasets"},{"id":"teleop","value":"12","label":"collected by teleoperation","figure":"by-kind"},{"id":"largest","value":"100,405 h","label":"largest: Egocentric-100K","figure":"largest"}],"findings":[{"id":"kinds","figure":"by-kind","text":"32 open datasets since 2023: 16 robot manipulation, 10 first-person video, 4 driving and 2 drones."},{"id":"largest","figure":"largest","text":"The largest by stated hours is Egocentric-100K (Build AI, 19 Dec 2025): 100,405 hours."},{"id":"teleop","figure":"datasets","text":"12 of the 32 were collected by people teleoperating robots, the work this site's teleoperation listings pay for."},{"id":"newest","figure":"datasets","text":"The newest release on record: EgoSuite-Open100K (26 Aug 2026). Lightwheel and Hugging Face opened the first 10,000 hours of a planned 100,000-hour egocentric human activity dataset recorded with head-mounted cameras (plus a wrist camera in the EgoPro subset), with hand and body pose annotations."}],"figures":[{"id":"datasets","title":"Open training datasets released since 2023","question":"Which public datasets teach robots, embodied AI, self-driving systems and drones, who released them, how big are they and how was the data collected?","columns":[{"key":"dataset","label":"Dataset","unit":"text"},{"key":"by","label":"Released by","unit":"text"},{"key":"domain","label":"Kind","unit":"text"},{"key":"method","label":"Collected by","unit":"text"},{"key":"hours","label":"Hours","unit":"count","secondary":true},{"key":"episodes","label":"Episodes","unit":"count","secondary":true},{"key":"date","label":"Released","unit":"text","secondary":true}],"rows":[{"key":"egosuite-open100k","source":{"label":"Lightwheel (Hugging Face blog)","url":"https://huggingface.co/blog/LightwheelAI/egosuite-open100k","date":"2026-08-26"},"cells":{"dataset":"EgoSuite-Open100K","by":"Lightwheel with Hugging Face","domain":"First-person video","method":"Wearable camera","hours":10000,"episodes":null,"date":"26 Aug 2026"}},{"key":"behavior-challenge-2026-demos","source":{"label":"Stanford Vision and Learning Lab","url":"https://behavior.stanford.edu/challenge/index.html","date":"2026-10-18"},"cells":{"dataset":"BEHAVIOR Challenge 2026 demonstrations","by":"Stanford Vision and Learning Lab (BEHAVIOR)","domain":"Robot manipulation","method":"Teleoperation","hours":1950,"episodes":20000,"date":"2 Jul 2026"}},{"key":"hiw-500","source":{"label":"BitRobot (Hugging Face)","url":"https://huggingface.co/datasets/BitRobot/HIW-500","date":"2026-10-18"},"cells":{"dataset":"HIW-500 (Humanoids-in-the-Wild 500)","by":"BitRobot Network with Hugging Face and Unitree Robotics","domain":"Robot manipulation","method":"Teleoperation","hours":500,"episodes":23000,"date":"25 Jun 2026"}},{"key":"abc-130k","source":{"label":"arXiv (UC Berkeley, MIT, Amazon FAR, XDOF, CMU)","url":"https://arxiv.org/abs/2606.27375","date":"2026-06-25"},"cells":{"dataset":"ABC-130K","by":"XDOF with UC Berkeley, MIT, Amazon FAR and Carnegie Mellon University","domain":"Robot manipulation","method":"Teleoperation","hours":3500,"episodes":130000,"date":"25 Jun 2026"}},{"key":"molmoact-2-bimanual-yam","source":{"label":"Ai2","url":"https://allenai.org/blog/molmoact2","date":"2026-05-05"},"cells":{"dataset":"MolmoAct 2-Bimanual YAM","by":"Ai2 (Allen Institute for AI)","domain":"Robot manipulation","method":"Not stated or other","hours":720,"episodes":32246,"date":"5 May 2026"}},{"key":"egocentric-100k","source":{"label":"Build AI (Hugging Face)","url":"https://huggingface.co/datasets/builddotai/Egocentric-100K","date":"2026-10-18"},"cells":{"dataset":"Egocentric-100K","by":"Build AI","domain":"First-person video","method":"Wearable camera","hours":100405,"episodes":null,"date":"19 Dec 2025"}},{"key":"robocoin","source":{"label":"arXiv (Beijing Academy of Artificial Intelligence)","url":"https://arxiv.org/abs/2511.17441","date":"2025-11-21"},"cells":{"dataset":"RoboCOIN","by":"Beijing Academy of Artificial Intelligence (BAAI) and partner institutions","domain":"Robot manipulation","method":"Teleoperation","hours":null,"episodes":180000,"date":"21 Nov 2025"}},{"key":"egocentric-10k","source":{"label":"Build AI (Hugging Face)","url":"https://huggingface.co/datasets/builddotai/Egocentric-10K","date":"2026-10-18"},"cells":{"dataset":"Egocentric-10K","by":"Build AI","domain":"First-person video","method":"Wearable camera","hours":10000,"episodes":null,"date":"11 Nov 2025"}},{"key":"wod-e2e","source":{"label":"arXiv (Waymo)","url":"https://arxiv.org/abs/2510.26125","date":"2025-10-30"},"cells":{"dataset":"Waymo Open Dataset for End-to-End Driving (WOD-E2E)","by":"Waymo","domain":"Driving","method":"Not stated or other","hours":12,"episodes":null,"date":"30 Oct 2025"}},{"key":"humanoid-everyday","source":{"label":"arXiv (USC, Toyota Research Institute)","url":"https://arxiv.org/abs/2510.08807","date":"2025-10-09"},"cells":{"dataset":"Humanoid Everyday","by":"University of Southern California and Toyota Research Institute","domain":"Robot manipulation","method":"Teleoperation","hours":null,"episodes":10300,"date":"9 Oct 2025"}},{"key":"galaxea-open-world-dataset","source":{"label":"arXiv (Galaxea AI)","url":"https://arxiv.org/html/2509.00576","date":"2025-08-30"},"cells":{"dataset":"Galaxea Open-World Dataset","by":"Galaxea AI","domain":"Robot manipulation","method":"Teleoperation","hours":500,"episodes":100000,"date":"30 Aug 2025"}},{"key":"molmoact-dataset","source":{"label":"Ai2","url":"https://allenai.org/blog/molmoact","date":"2025-08-12"},"cells":{"dataset":"MolmoAct Dataset","by":"Ai2 (Allen Institute for AI)","domain":"Robot manipulation","method":"Not stated or other","hours":22,"episodes":10000,"date":"12 Aug 2025"}},{"key":"lerobot-community-datasets","source":{"label":"Hugging Face","url":"https://huggingface.co/blog/smolvla","date":"2025-06-03"},"cells":{"dataset":"LeRobot community datasets (SmolVLA pretraining set)","by":"Hugging Face (LeRobot) and community contributors","domain":"Robot manipulation","method":"Mixed","hours":null,"episodes":null,"date":"3 Jun 2025"}},{"key":"uav-flow","source":{"label":"arXiv (Beihang University, NUS)","url":"https://arxiv.org/html/2505.15725","date":"2025-05-21"},"cells":{"dataset":"UAV-Flow","by":"Beihang University and National University of Singapore","domain":"Drones","method":"Teleoperation","hours":null,"episodes":30000,"date":"21 May 2025"}},{"key":"egodex","source":{"label":"arXiv (Apple)","url":"https://arxiv.org/abs/2505.11709","date":"2025-05-16"},"cells":{"dataset":"EgoDex","by":"Apple","domain":"First-person video","method":"Wearable camera","hours":829,"episodes":338000,"date":"16 May 2025"}},{"key":"fourier-actionnet","source":{"label":"Fourier (ActionNet project page)","url":"https://action-net.org/","date":"2026-10-18"},"cells":{"dataset":"Fourier ActionNet","by":"Fourier","domain":"Robot manipulation","method":"Teleoperation","hours":140,"episodes":30000,"date":"18 Mar 2025"}},{"key":"hd-epic","source":{"label":"arXiv (University of Bristol and partners)","url":"https://arxiv.org/abs/2502.04144","date":"2025-02-06"},"cells":{"dataset":"HD-EPIC","by":"University of Bristol, Leiden University, Singapore Management University and University of Bath (EPIC-KITCHENS team)","domain":"First-person video","method":"Wearable camera","hours":41,"episodes":null,"date":"6 Feb 2025"}},{"key":"agibot-world","source":{"label":"arXiv (AgiBot World Colosseo team)","url":"https://arxiv.org/html/2503.06669","date":"2025-03-09"},"cells":{"dataset":"AgiBot World","by":"AgiBot","domain":"Robot manipulation","method":"Teleoperation","hours":2976.4,"episodes":1001552,"date":"30 Dec 2024"}},{"key":"robomind","source":{"label":"arXiv (Beijing Innovation Center of Humanoid Robotics, Peking University)","url":"https://arxiv.org/abs/2412.13877","date":"2024-12-18"},"cells":{"dataset":"RoboMIND","by":"Beijing Innovation Center of Humanoid Robotics (X-Humanoid), Peking University and Beijing Academy of Artificial Intelligence","domain":"Robot manipulation","method":"Teleoperation","hours":305.5,"episodes":107000,"date":"18 Dec 2024"}},{"key":"man-truckscenes","source":{"label":"arXiv (MAN Truck & Bus, TUM)","url":"https://arxiv.org/abs/2407.07462","date":"2024-07-10"},"cells":{"dataset":"MAN TruckScenes","by":"MAN Truck & Bus with the Technical University of Munich","domain":"Driving","method":"Not stated or other","hours":null,"episodes":null,"date":"10 Jul 2024"}},{"key":"citynav","source":{"label":"arXiv","url":"https://arxiv.org/abs/2406.14240","date":"2024-06-20"},"cells":{"dataset":"CityNav","by":"Tokyo Institute of Technology, The University of Tokyo, Kyoto University and partners","domain":"Drones","method":"Not stated or other","hours":null,"episodes":32637,"date":"20 Jun 2024"}},{"key":"nymeria","source":{"label":"arXiv (Meta Reality Labs Research)","url":"https://arxiv.org/abs/2406.09905","date":"2024-06-14"},"cells":{"dataset":"Nymeria","by":"Meta (Reality Labs Research, Project Aria)","domain":"First-person video","method":"Wearable camera","hours":300,"episodes":null,"date":"14 Jun 2024"}},{"key":"hot3d","source":{"label":"arXiv (Meta)","url":"https://arxiv.org/abs/2406.09598","date":"2024-06-13"},"cells":{"dataset":"HOT3D","by":"Meta (Reality Labs)","domain":"First-person video","method":"Wearable camera","hours":null,"episodes":null,"date":"13 Jun 2024"}},{"key":"droid","source":{"label":"arXiv (DROID collaboration)","url":"https://arxiv.org/abs/2403.12945","date":"2024-03-19"},"cells":{"dataset":"DROID (Distributed Robot Interaction Dataset)","by":"DROID collaboration (13 institutions)","domain":"Robot manipulation","method":"Teleoperation","hours":350,"episodes":76000,"date":"19 Mar 2024"}},{"key":"aria-everyday-activities","source":{"label":"arXiv (Meta)","url":"https://arxiv.org/abs/2402.13349","date":"2024-02-20"},"cells":{"dataset":"Aria Everyday Activities (AEA)","by":"Meta (Project Aria)","domain":"First-person video","method":"Wearable camera","hours":7.5,"episodes":null,"date":"20 Feb 2024"}},{"key":"ego-exo4d","source":{"label":"arXiv (Meta FAIR and university partners)","url":"https://arxiv.org/abs/2311.18259","date":"2023-11-30"},"cells":{"dataset":"Ego-Exo4D","by":"Meta (FAIR and Project Aria) with 14 university partners","domain":"First-person video","method":"Mixed","hours":1286,"episodes":null,"date":"30 Nov 2023"}},{"key":"open-x-embodiment","source":{"label":"arXiv (Open X-Embodiment Collaboration)","url":"https://arxiv.org/abs/2310.08864","date":"2023-10-13"},"cells":{"dataset":"Open X-Embodiment","by":"Google DeepMind and the Open X-Embodiment Collaboration (21 institutions)","domain":"Robot manipulation","method":"Mixed","hours":null,"episodes":1000000,"date":"3 Oct 2023"}},{"key":"bridgedata-v2","source":{"label":"arXiv (UC Berkeley, Stanford)","url":"https://arxiv.org/abs/2308.12952","date":"2023-08-24"},"cells":{"dataset":"BridgeData V2","by":"UC Berkeley (RAIL) with Stanford and collaborators","domain":"Robot manipulation","method":"Mixed","hours":null,"episodes":60096,"date":"24 Aug 2023"}},{"key":"rh20t","source":{"label":"arXiv (Shanghai Jiao Tong University)","url":"https://arxiv.org/abs/2307.00595","date":"2023-07-02"},"cells":{"dataset":"RH20T","by":"Shanghai Jiao Tong University","domain":"Robot manipulation","method":"Teleoperation","hours":null,"episodes":110000,"date":"2 Jul 2023"}},{"key":"zenseact-open-dataset","source":{"label":"arXiv (Zenseact)","url":"https://arxiv.org/abs/2305.02008","date":"2023-05-03"},"cells":{"dataset":"Zenseact Open Dataset (ZOD)","by":"Zenseact","domain":"Driving","method":"Not stated or other","hours":null,"episodes":null,"date":"3 May 2023"}},{"key":"ego4d","source":{"label":"Ego4D","url":"https://ego4d-data.org/docs/updates/","date":"2026-10-18"},"cells":{"dataset":"Ego4D (version 2)","by":"Meta (FAIR) and the Ego4D university consortium","domain":"First-person video","method":"Wearable camera","hours":3670,"episodes":null,"date":"1 Feb 2023"}},{"key":"argoverse-2","source":{"label":"arXiv (Argo AI)","url":"https://arxiv.org/abs/2301.00493","date":"2023-01-02"},"cells":{"dataset":"Argoverse 2","by":"Argo AI","domain":"Driving","method":"Not stated or other","hours":null,"episodes":null,"date":"2 Jan 2023"}}],"sample":"32 datasets released from 2023 to 18 Oct 2026, each linked to its paper or release page (reviewed 18 Oct 2026).","notes":["Hours and episodes are as each release states them; a blank means the release gives no figure. Episodes, trajectories and demonstrations are counted as one."]},{"id":"by-kind","title":"Datasets by what they teach","question":"How many open datasets are there for each kind of robot learning, and how was their data gathered?","columns":[{"key":"domain","label":"Kind","unit":"text"},{"key":"datasets","label":"Datasets","unit":"count","bar":true},{"key":"teleop","label":"Teleoperated","unit":"count"},{"key":"wearable","label":"Wearable camera","unit":"count"},{"key":"largest","label":"Largest by stated hours","unit":"text","secondary":true}],"rows":[{"key":"robot","cells":{"domain":"Robot manipulation","datasets":16,"teleop":11,"wearable":0,"largest":"ABC-130K (3,500 h)"}},{"key":"egocentric","cells":{"domain":"First-person video","datasets":10,"teleop":0,"wearable":9,"largest":"Egocentric-100K (100,405 h)"}},{"key":"driving","cells":{"domain":"Driving","datasets":4,"teleop":0,"wearable":0,"largest":"Waymo Open Dataset for End-to-End Driving (WOD-E2E) (12 h)"}},{"key":"drone","cells":{"domain":"Drones","datasets":2,"teleop":1,"wearable":0,"largest":"n/a"}}],"sample":"32 datasets.","notes":["Collected by: how the release says the data was gathered. Teleoperation means people steering robots remotely or in VR; wearable camera means people filming their own hands and tasks."]},{"id":"largest","title":"The largest datasets by hours of recorded data","question":"Which open datasets hold the most hours of human-made robot training data?","columns":[{"key":"dataset","label":"Dataset","unit":"text"},{"key":"hours","label":"Hours","unit":"count","bar":true},{"key":"method","label":"Collected by","unit":"text","secondary":true},{"key":"contributors","label":"Who recorded it","unit":"text","secondary":true}],"rows":[{"key":"egocentric-100k","source":{"label":"Build AI (Hugging Face)","url":"https://huggingface.co/datasets/builddotai/Egocentric-100K","date":"2026-10-18"},"cells":{"dataset":"Egocentric-100K","hours":100405,"method":"Wearable camera","contributors":"14,228 factory workers (as reported from the dataset's technical details)"}},{"key":"egosuite-open100k","source":{"label":"Lightwheel (Hugging Face blog)","url":"https://huggingface.co/blog/LightwheelAI/egosuite-open100k","date":"2026-08-26"},"cells":{"dataset":"EgoSuite-Open100K","hours":10000,"method":"Wearable camera","contributors":"A globally distributed workforce of tens of thousands of collectors"}},{"key":"egocentric-10k","source":{"label":"Build AI (Hugging Face)","url":"https://huggingface.co/datasets/builddotai/Egocentric-10K","date":"2026-10-18"},"cells":{"dataset":"Egocentric-10K","hours":10000,"method":"Wearable camera","contributors":"2,153 factory workers (as reported), in factories numbered 001 to 085 on the dataset card"}},{"key":"ego4d","source":{"label":"Ego4D","url":"https://ego4d-data.org/docs/updates/","date":"2026-10-18"},"cells":{"dataset":"Ego4D (version 2)","hours":3670,"method":"Wearable camera","contributors":"923 unique participants in 74 locations in 9 countries; 88 researchers in an international consortium"}},{"key":"abc-130k","source":{"label":"arXiv (UC Berkeley, MIT, Amazon FAR, XDOF, CMU)","url":"https://arxiv.org/abs/2606.27375","date":"2026-06-25"},"cells":{"dataset":"ABC-130K","hours":3500,"method":"Teleoperation","contributors":"n/a"}},{"key":"agibot-world","source":{"label":"arXiv (AgiBot World Colosseo team)","url":"https://arxiv.org/html/2503.06669","date":"2025-03-09"},"cells":{"dataset":"AgiBot World","hours":2976.4,"method":"Teleoperation","contributors":"Skilled teleoperators working more than 100 homogeneous robots in a data collection facility of over 4,000 square meters"}},{"key":"behavior-challenge-2026-demos","source":{"label":"Stanford Vision and Learning Lab","url":"https://behavior.stanford.edu/challenge/index.html","date":"2026-10-18"},"cells":{"dataset":"BEHAVIOR Challenge 2026 demonstrations","hours":1950,"method":"Teleoperation","contributors":"Teleoperation data in simulation provided by Simovation"}},{"key":"ego-exo4d","source":{"label":"arXiv (Meta FAIR and university partners)","url":"https://arxiv.org/abs/2311.18259","date":"2023-11-30"},"cells":{"dataset":"Ego-Exo4D","hours":1286,"method":"Mixed","contributors":"740 participants (camera wearers) in 13 cities; Meta's FAIR, Project Aria and 14 university partners"}},{"key":"egodex","source":{"label":"arXiv (Apple)","url":"https://arxiv.org/abs/2505.11709","date":"2025-05-16"},"cells":{"dataset":"EgoDex","hours":829,"method":"Wearable camera","contributors":"n/a"}},{"key":"molmoact-2-bimanual-yam","source":{"label":"Ai2","url":"https://allenai.org/blog/molmoact2","date":"2026-05-05"},"cells":{"dataset":"MolmoAct 2-Bimanual YAM","hours":720,"method":"Not stated or other","contributors":"Curated with support from Cortex AI"}},{"key":"hiw-500","source":{"label":"BitRobot (Hugging Face)","url":"https://huggingface.co/datasets/BitRobot/HIW-500","date":"2026-10-18"},"cells":{"dataset":"HIW-500 (Humanoids-in-the-Wild 500)","hours":500,"method":"Teleoperation","contributors":"Human teleoperators in 12 homes across Southeast Asia"}},{"key":"galaxea-open-world-dataset","source":{"label":"arXiv (Galaxea AI)","url":"https://arxiv.org/html/2509.00576","date":"2025-08-30"},"cells":{"dataset":"Galaxea Open-World Dataset","hours":500,"method":"Teleoperation","contributors":"n/a"}},{"key":"droid","source":{"label":"arXiv (DROID collaboration)","url":"https://arxiv.org/abs/2403.12945","date":"2024-03-19"},"cells":{"dataset":"DROID (Distributed Robot Interaction Dataset)","hours":350,"method":"Teleoperation","contributors":"50 data collectors in North America, Asia and Europe at 13 institutions, over 12 months"}},{"key":"robomind","source":{"label":"arXiv (Beijing Innovation Center of Humanoid Robotics, Peking University)","url":"https://arxiv.org/abs/2412.13877","date":"2024-12-18"},"cells":{"dataset":"RoboMIND","hours":305.5,"method":"Teleoperation","contributors":"n/a"}},{"key":"nymeria","source":{"label":"arXiv (Meta Reality Labs Research)","url":"https://arxiv.org/abs/2406.09905","date":"2024-06-14"},"cells":{"dataset":"Nymeria","hours":300,"method":"Wearable camera","contributors":"264 participants across 50 locations"}},{"key":"fourier-actionnet","source":{"label":"Fourier (ActionNet project page)","url":"https://action-net.org/","date":"2026-10-18"},"cells":{"dataset":"Fourier ActionNet","hours":140,"method":"Teleoperation","contributors":"n/a"}},{"key":"hd-epic","source":{"label":"arXiv (University of Bristol and partners)","url":"https://arxiv.org/abs/2502.04144","date":"2025-02-06"},"cells":{"dataset":"HD-EPIC","hours":41,"method":"Wearable camera","contributors":"9 participants recording in their own kitchens for at least 3 consecutive days"}},{"key":"molmoact-dataset","source":{"label":"Ai2","url":"https://allenai.org/blog/molmoact","date":"2025-08-12"},"cells":{"dataset":"MolmoAct Dataset","hours":22,"method":"Not stated or other","contributors":"Collected in-house by Ai2 over about 3 months"}},{"key":"wod-e2e","source":{"label":"arXiv (Waymo)","url":"https://arxiv.org/abs/2510.26125","date":"2025-10-30"},"cells":{"dataset":"Waymo Open Dataset for End-to-End Driving (WOD-E2E)","hours":12,"method":"Not stated or other","contributors":"n/a"}},{"key":"aria-everyday-activities","source":{"label":"arXiv (Meta)","url":"https://arxiv.org/abs/2402.13349","date":"2024-02-20"},"cells":{"dataset":"Aria Everyday Activities (AEA)","hours":7.5,"method":"Wearable camera","contributors":"Multiple wearers (1 to 2 per sequence) in five geographically diverse indoor locations"}}],"sample":"20 of 32 datasets state their hours.","notes":["Hours are not totalled across datasets: releases count them differently (raw video, cleaned episodes, approved data)."]},{"id":"licences","title":"Size, contributors and licence of each dataset","question":"What does each dataset contain, who recorded it and under what licence can it be used?","columns":[{"key":"dataset","label":"Dataset","unit":"text"},{"key":"size","label":"Size as stated","unit":"text"},{"key":"licence","label":"Licence","unit":"text","secondary":true}],"rows":[{"key":"egosuite-open100k-size","source":{"label":"Lightwheel (Hugging Face blog)","url":"https://huggingface.co/blog/LightwheelAI/egosuite-open100k","date":"2026-08-26"},"cells":{"dataset":"EgoSuite-Open100K","size":"100,000 hours planned; first 10,000 hours available; 15,000+ tasks and 15,000+ real-world collection scenes","licence":"commercial-training-no-resale-v1.0 (academic research and commercial training)"}},{"key":"behavior-challenge-2026-demos-size","source":{"label":"Stanford Vision and Learning Lab","url":"https://behavior.stanford.edu/challenge/index.html","date":"2026-10-18"},"cells":{"dataset":"BEHAVIOR Challenge 2026 demonstrations","size":"20,000 human teleoperation demos, 1,950 hours in total (in simulation)","licence":"MIT"}},{"key":"hiw-500-size","source":{"label":"BitRobot (Hugging Face)","url":"https://huggingface.co/datasets/BitRobot/HIW-500","date":"2026-10-18"},"cells":{"dataset":"HIW-500 (Humanoids-in-the-Wild 500)","size":"500+ hours, 23K+ episodes and 10+ TB of raw data from 12 real homes","licence":"CC BY 4.0"}},{"key":"abc-130k-size","source":{"label":"arXiv (UC Berkeley, MIT, Amazon FAR, XDOF, CMU)","url":"https://arxiv.org/abs/2606.27375","date":"2026-06-25"},"cells":{"dataset":"ABC-130K","size":"3,500 hours over 130K episodes across 195 tasks (project page: 134,806 episodes, 3,553 hours), plus 400 hours of sim-teleop data","licence":"Apache 2.0"}},{"key":"molmoact-2-bimanual-yam-size","source":{"label":"Ai2","url":"https://allenai.org/blog/molmoact2","date":"2026-05-05"},"cells":{"dataset":"MolmoAct 2-Bimanual YAM","size":"over 720 hours of training demonstrations (the blog also calls it a 700-hour collection); 32,246 episodes on the Hugging Face card","licence":"Apache 2.0"}},{"key":"egocentric-100k-size","source":{"label":"Build AI (Hugging Face)","url":"https://huggingface.co/datasets/builddotai/Egocentric-100K","date":"2026-10-18"},"cells":{"dataset":"Egocentric-100K","size":"100,405 hours, 10.8 billion frames, 2,010,759 clips, 24.79 TB","licence":"Apache 2.0"}},{"key":"robocoin-size","source":{"label":"arXiv (Beijing Academy of Artificial Intelligence)","url":"https://arxiv.org/abs/2511.17441","date":"2025-11-21"},"cells":{"dataset":"RoboCOIN","size":"over 180,000 demonstrations from 15 robotic platforms, 421 bimanual tasks, 432 objects, 16 environments","licence":"Not stated on the release page"}},{"key":"egocentric-10k-size","source":{"label":"Build AI (Hugging Face)","url":"https://huggingface.co/datasets/builddotai/Egocentric-10K","date":"2026-10-18"},"cells":{"dataset":"Egocentric-10K","size":"10,000 hours, 1.08 billion frames, 192,900 clips, 16.4 TB","licence":"Apache 2.0"}},{"key":"wod-e2e-size","source":{"label":"arXiv (Waymo)","url":"https://arxiv.org/abs/2510.26125","date":"2025-10-30"},"cells":{"dataset":"Waymo Open Dataset for End-to-End Driving (WOD-E2E)","size":"4,021 driving segments (approximately 12 hours) of 20 seconds each, curated for long-tail scenarios","licence":"Waymo Dataset License Agreement for Non-Commercial Use"}},{"key":"humanoid-everyday-size","source":{"label":"arXiv (USC, Toyota Research Institute)","url":"https://arxiv.org/abs/2510.08807","date":"2025-10-09"},"cells":{"dataset":"Humanoid Everyday","size":"10.3k trajectories and over 3 million frames across 260 tasks in 7 categories","licence":"Not stated on the release page"}},{"key":"galaxea-open-world-dataset-size","source":{"label":"arXiv (Galaxea AI)","url":"https://arxiv.org/html/2509.00576","date":"2025-08-30"},"cells":{"dataset":"Galaxea Open-World Dataset","size":"500 hours, 100K demonstration trajectories, more than 150 tasks across 50 real-world scenes","licence":"CC BY-NC-SA 4.0"}},{"key":"molmoact-dataset-size","source":{"label":"Ai2","url":"https://allenai.org/blog/molmoact","date":"2025-08-12"},"cells":{"dataset":"MolmoAct Dataset","size":"~10,000 distinct robot episodes (the MolmoAct 2 announcement describes 22 hours, about 10.6K trajectories)","licence":"CC BY 4.0"}},{"key":"lerobot-community-datasets-size","source":{"label":"Hugging Face","url":"https://huggingface.co/blog/smolvla","date":"2025-06-03"},"cells":{"dataset":"LeRobot community datasets (SmolVLA pretraining set)","size":"487 curated community datasets focused on the SO100 robot arm, standardized at 30 FPS, around 10 million frames (the SmolVLA paper counts 481 datasets)","licence":"Varies by dataset (compatibly licensed, open-source community datasets)"}},{"key":"uav-flow-size","source":{"label":"arXiv (Beihang University, NUS)","url":"https://arxiv.org/html/2505.15725","date":"2025-05-21"},"cells":{"dataset":"UAV-Flow","size":"real-world dataset with 30k flight trajectories, plus a simulation dataset","licence":"Not stated on the release page"}},{"key":"egodex-size","source":{"label":"arXiv (Apple)","url":"https://arxiv.org/abs/2505.11709","date":"2025-05-16"},"cells":{"dataset":"EgoDex","size":"829 hours of 1080p, 30 Hz egocentric video with 338,000 episodes across 194 tasks (90 million frames)","licence":"CC BY-NC-ND"}},{"key":"fourier-actionnet-size","source":{"label":"Fourier (ActionNet project page)","url":"https://action-net.org/","date":"2026-10-18"},"cells":{"dataset":"Fourier ActionNet","size":"over 30K teleoperated trajectories or around 140 hours of interaction data","licence":"CC BY-NC-SA 4.0 (dataset); Apache 2.0 (training and evaluation tools)"}},{"key":"hd-epic-size","source":{"label":"arXiv (University of Bristol and partners)","url":"https://arxiv.org/abs/2502.04144","date":"2025-02-06"},"cells":{"dataset":"HD-EPIC","size":"41 hours of video in 9 kitchens, 69 recipes, 59K fine-grained actions, 51K audio events, 20K object movements","licence":"Not stated on the release page"}},{"key":"agibot-world-size","source":{"label":"arXiv (AgiBot World Colosseo team)","url":"https://arxiv.org/html/2503.06669","date":"2025-03-09"},"cells":{"dataset":"AgiBot World","size":"1,001,552 trajectories totalling 2,976.4 hours, 217 tasks, 87 skills and 106 scenes (latest version per the paper); the GitHub Beta release lists 1,003,672 trajectories (~43.8 TB)","licence":"CC BY-NC-SA 4.0"}},{"key":"robomind-size","source":{"label":"arXiv (Beijing Innovation Center of Humanoid Robotics, Peking University)","url":"https://arxiv.org/abs/2412.13877","date":"2024-12-18"},"cells":{"dataset":"RoboMIND","size":"107k demonstration trajectories (305.5 hours) across 479 tasks and 96 object classes on 4 robot embodiments, plus 5k real-world failure demonstrations","licence":"Apache 2.0 (Hugging Face listing)"}},{"key":"man-truckscenes-size","source":{"label":"arXiv (MAN Truck & Bus, TUM)","url":"https://arxiv.org/abs/2407.07462","date":"2024-07-10"},"cells":{"dataset":"MAN TruckScenes","size":"more than 740 scenes of 20 s each, with 4 cameras, 6 lidar and 6 radar sensors","licence":"CC BY-NC-SA 4.0"}},{"key":"citynav-size","source":{"label":"arXiv","url":"https://arxiv.org/abs/2406.14240","date":"2024-06-20"},"cells":{"dataset":"CityNav","size":"32,637 human demonstration trajectories paired with language descriptions, over 4.65 km2 of Cambridge and Birmingham","licence":"Not stated on the release page"}},{"key":"nymeria-size","source":{"label":"arXiv (Meta Reality Labs Research)","url":"https://arxiv.org/abs/2406.09905","date":"2024-06-14"},"cells":{"dataset":"Nymeria","size":"300 hours of daily activities from 264 participants across 50 locations, over 399 km travelled","licence":"Nymeria dataset license (Meta's own terms, in the repository)"}},{"key":"hot3d-size","source":{"label":"arXiv (Meta)","url":"https://arxiv.org/abs/2406.09598","date":"2024-06-13"},"cells":{"dataset":"HOT3D","size":"over 833 minutes (more than 3.7M images) of multi-view image streams, 19 subjects, 33 rigid objects","licence":"HOT3D license agreement (Meta)"}},{"key":"droid-size","source":{"label":"arXiv (DROID collaboration)","url":"https://arxiv.org/abs/2403.12945","date":"2024-03-19"},"cells":{"dataset":"DROID (Distributed Robot Interaction Dataset)","size":"76k demonstration trajectories or 350 hours of interaction data, across 564 scenes and 84 tasks (the project page says 86 tasks)","licence":"CC BY 4.0"}},{"key":"aria-everyday-activities-size","source":{"label":"arXiv (Meta)","url":"https://arxiv.org/abs/2402.13349","date":"2024-02-20"},"cells":{"dataset":"Aria Everyday Activities (AEA)","size":"143 daily activity sequences, over 7.5 accumulated hours, recorded in 5 locations","licence":"Not stated on the release page"}},{"key":"ego-exo4d-size","source":{"label":"arXiv (Meta FAIR and university partners)","url":"https://arxiv.org/abs/2311.18259","date":"2023-11-30"},"cells":{"dataset":"Ego-Exo4D","size":"1,286 hours of video combined (1,286.3 ego and exo hours) from 740 participants in 13 cities and 123 scene contexts","licence":"Ego-Exo4D license agreement (must be signed and approved before download)"}},{"key":"open-x-embodiment-size","source":{"label":"arXiv (Open X-Embodiment Collaboration)","url":"https://arxiv.org/abs/2310.08864","date":"2023-10-13"},"cells":{"dataset":"Open X-Embodiment","size":"1M+ real robot trajectories spanning 22 robot embodiments, pooled from 60 existing robot datasets; 527 skills (160,266 tasks)","licence":"CC BY 4.0 for data and materials, Apache 2.0 for software"}},{"key":"bridgedata-v2-size","source":{"label":"arXiv (UC Berkeley, Stanford)","url":"https://arxiv.org/abs/2308.12952","date":"2023-08-24"},"cells":{"dataset":"BridgeData V2","size":"60,096 trajectories across 24 environments on a WidowX 250 arm: 50,365 teleoperated demonstrations and 9,731 rollouts from a scripted pick-and-place policy","licence":"CC BY 4.0"}},{"key":"rh20t-size","source":{"label":"arXiv (Shanghai Jiao Tong University)","url":"https://arxiv.org/abs/2307.00595","date":"2023-07-02"},"cells":{"dataset":"RH20T","size":"over 110,000 contact-rich robot manipulation sequences, each paired with a human demonstration video; tasks drawn from RLBench (48), MetaWorld (29) and 70 self-proposed tasks","licence":"Mixed: RH20T-C subset CC BY-SA 4.0, RH20T-NC subset CC BY-NC 4.0"}},{"key":"zenseact-open-dataset-size","source":{"label":"arXiv (Zenseact)","url":"https://arxiv.org/abs/2305.02008","date":"2023-05-03"},"cells":{"dataset":"Zenseact Open Dataset (ZOD)","size":"100k curated camera frames, 1,473 sequences of 20 seconds and 29 drives of a few minutes, collected over two years in 14 European countries","licence":"CC BY-SA 4.0"}},{"key":"ego4d-size","source":{"label":"Ego4D","url":"https://ego4d-data.org/docs/updates/","date":"2026-10-18"},"cells":{"dataset":"Ego4D (version 2)","size":"3,670 hours of daily-life video from 923 participants in 74 locations across 9 countries","licence":"Ego4D license agreement (must be accepted and approved before download)"}},{"key":"argoverse-2-size","source":{"label":"arXiv (Argo AI)","url":"https://arxiv.org/abs/2301.00493","date":"2023-01-02"},"cells":{"dataset":"Argoverse 2","size":"Sensor Dataset: 1,000 annotated sequences; Lidar Dataset: 20,000 sequences; Motion Forecasting Dataset: 250,000 scenarios; six US cities","licence":"CC BY-NC-SA 4.0"}}],"sample":"All 32 datasets.","notes":["A licence is summarised as the release page gives it; read the licence itself before using a dataset."]}]}