{"apiVersion":"v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/","scope":"Normalized factual records. Original source material, photographs, code, models and datasets retain their own licences."},"attribution":{"name":"humanoidversus.com","url":"https://humanoidversus.com/data","required":true,"text":"Data: humanoidversus.com. Link to the canonical record or /data, retain source references and verification dates, and indicate any changes."},"lastModified":"2026-10-13","data":[{"slug":"robocasa365","name":"RoboCasa365","publisher":{"name":"RoboCasa Team (University of Texas at Austin)","company":null},"released":"2026-02-18","summary":"Simulation datasets for pretraining and evaluating generalist kitchen-manipulation policies in RoboCasa365. The paper uses a Franka Panda arm on an Omron mobile base and describes human teleoperation and MimicGen-generated demonstrations. The datasets are licensed CC BY 4.0 and the current download documentation distributes LeRobot files.","license":"CC BY 4.0","commercialUse":true,"size":{"episodes":null,"hours":null,"trajectories":null,"storageTB":null,"note":"Release notes publish 2,200+ hours. The RoboCasa365 paper reports 612 human hours and 1,615 synthetic hours, with human pretraining inconsistently given as 404 and 411 hours; no combined duration is stored. It publishes 30k human pretraining demonstrations and 25k target demonstrations. No computed total or storage conversion is stored."},"robots":[{"robot":null,"name":"Franka Panda with Omron mobile base (simulated)"}],"modalities":["rgb","proprioception","language"],"format":["LeRobot"],"realOrSim":"sim","humanoid":false,"links":{"homepage":"https://robocasa.ai/","download":"https://github.com/robocasa/robocasa/blob/main/docs/datasets/using_datasets.md","paper":"https://arxiv.org/abs/2603.04356"},"sources":[{"id":"s1","url":"https://robocasa.ai/","title":"RoboCasa365 project page","publisher":"RoboCasa team","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://github.com/robocasa/robocasa","title":"robocasa repository README (license, release notes)","publisher":"RoboCasa team","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s3","url":"https://arxiv.org/abs/2603.04356","title":"RoboCasa365: A Large-Scale Simulation Framework for Training and Benchmarking Generalist Robots","publisher":"arXiv","date":"2026-03-04","type":"paper","accessed":"2026-10-13"},{"id":"s4","url":"https://arxiv.org/html/2603.04356","title":"RoboCasa365 paper: robot embodiment and dataset statistics","publisher":"arXiv","date":null,"type":"paper","accessed":"2026-10-13"},{"id":"s5","url":"https://github.com/robocasa/robocasa/blob/main/docs/datasets/using_datasets.md","title":"RoboCasa365 current dataset download and format documentation","publisher":"RoboCasa Team","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s6","url":"https://creativecommons.org/licenses/by/4.0/","title":"Creative Commons Attribution 4.0 International","publisher":"Creative Commons","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/robocasa365"},{"slug":"galaxea-open-world","name":"Galaxea Open-World Dataset","publisher":{"name":"Galaxea (OpenGalaxea)","company":"galaxea"},"released":"2025-09-09","summary":"Real-world mobile manipulation dataset of over 500 hours recorded on Galaxea R1-Lite robots in residential, kitchen, retail and office settings, with subtask language annotations. The CC BY-NC-SA 4.0 license bars commercial use.","license":"CC BY-NC-SA 4.0","commercialUse":false,"size":{"episodes":null,"hours":null,"trajectories":null,"storageTB":2.87,"note":"Card publishes 500+ hours, not an exact duration, and 227 task-level archives. Hugging Face lists 2.87 TB total repository files. R1-Lite is not linked to the catalog R1 slug because the collection generation is unspecified. Official repository dates the original data release to 2025-09-09 and LeRobot availability to 2025-10-07."},"robots":[{"robot":null,"name":"Galaxea R1-Lite"}],"modalities":["rgb","proprioception","imu","language"],"format":["RLDS","LeRobot","Parquet","MP4"],"realOrSim":"real","humanoid":true,"links":{"homepage":"https://huggingface.co/datasets/OpenGalaxea/Galaxea-Open-World-Dataset","download":"https://huggingface.co/datasets/OpenGalaxea/Galaxea-Open-World-Dataset","paper":"https://arxiv.org/abs/2509.00576"},"sources":[{"id":"s1","url":"https://huggingface.co/datasets/OpenGalaxea/Galaxea-Open-World-Dataset","title":"Galaxea-Open-World-Dataset dataset card","publisher":"OpenGalaxea","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://arxiv.org/abs/2509.00576","title":"Galaxea Open-World Dataset and G0 Dual-System VLA Model","publisher":"arXiv","date":"2025-08-30","type":"paper","accessed":"2026-10-13"},{"id":"s3","url":"https://github.com/OpenGalaxea/GalaxeaVLA","title":"GalaxeaVLA official release log and dataset formats","publisher":"Galaxea","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s4","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/","title":"Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International","publisher":"Creative Commons","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/galaxea-open-world"},{"slug":"agibot-world","name":"AgiBot World (Beta)","publisher":{"name":"AgiBot","company":"agibot"},"released":"2025-03-01","summary":"AgiBot's complete dataset (the Beta release) of real-robot manipulation trajectories recorded on AgiBot G1 platforms, described as over one million trajectories from 100 robots. The CC BY-NC-SA 4.0 license bars commercial use.","license":"CC BY-NC-SA 4.0","commercialUse":false,"size":{"episodes":null,"hours":2976.4,"trajectories":1003672,"storageTB":48.1,"note":"GitHub publishes 1,003,672 trajectories and ~43.8T; the Beta card publishes 2,976.4 hours and Hugging Face lists 48.1 TB total repository files. The storage figures differ and their relationship is not explained. Native files include HDF5 states, PNG depth and camera videos; LeRobot conversion scripts are provided. The card says action_text/description annotations are not yet available."},"robots":[{"robot":"agibot-genie-g1","name":"AgiBot G1"}],"modalities":["rgb","depth","proprioception"],"format":["HDF5","MP4","PNG","JSON"],"realOrSim":"real","humanoid":true,"links":{"homepage":"https://github.com/OpenDriveLab/AgiBot-World","download":"https://huggingface.co/datasets/agibot-world/AgiBotWorld-Beta","paper":"https://arxiv.org/abs/2503.06669"},"sources":[{"id":"s1","url":"https://github.com/OpenDriveLab/AgiBot-World","title":"AgiBot World repository README","publisher":"OpenDriveLab / AgiBot","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://huggingface.co/datasets/agibot-world/AgiBotWorld-Beta","title":"AgiBotWorld-Beta dataset card","publisher":"AgiBot","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s3","url":"https://arxiv.org/abs/2503.06669","title":"AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems","publisher":"arXiv","date":"2025-03-09","type":"paper","accessed":"2026-10-13"},{"id":"s4","url":"https://opendrivelab.com/AgiBot-World/","title":"AgiBot World Colosseo official platform description","publisher":"OpenDriveLab","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s5","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/","title":"Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International","publisher":"Creative Commons","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/agibot-world"},{"slug":"agibot-world-alpha","name":"AgiBot World Alpha","publisher":{"name":"AgiBot","company":"agibot"},"released":"2024-12-30","summary":"Curated Alpha subset of AgiBot World, the first public release, recorded on AgiBot G1 platforms. The CC BY-NC-SA 4.0 license bars commercial use.","license":"CC BY-NC-SA 4.0","commercialUse":false,"size":{"episodes":null,"hours":595.31,"trajectories":null,"storageTB":9.66,"note":"Updated Alpha notice publishes 595.31 hours but no exact updated trajectory count. The card’s feature bullet says 100,000+ trajectories and 300 hours; GitHub lists 92,214 trajectories and ~8.5T. These statements conflict, so trajectory count is null. Hugging Face lists 9.66 TB total repository files. LeRobot conversion scripts are provided."},"robots":[{"robot":"agibot-genie-g1","name":"AgiBot G1"}],"modalities":["rgb","depth","proprioception","language"],"format":["HDF5","MP4","PNG","JSON"],"realOrSim":"real","humanoid":true,"links":{"homepage":"https://huggingface.co/datasets/agibot-world/AgiBotWorld-Alpha","download":"https://huggingface.co/datasets/agibot-world/AgiBotWorld-Alpha","paper":"https://arxiv.org/abs/2503.06669"},"sources":[{"id":"s1","url":"https://huggingface.co/datasets/agibot-world/AgiBotWorld-Alpha","title":"AgiBotWorld-Alpha dataset card","publisher":"AgiBot","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://github.com/OpenDriveLab/AgiBot-World","title":"AgiBot World repository README","publisher":"OpenDriveLab / AgiBot","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s3","url":"https://arxiv.org/abs/2503.06669","title":"AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems","publisher":"arXiv","date":"2025-03-09","type":"paper","accessed":"2026-10-13"},{"id":"s4","url":"https://opendrivelab.com/AgiBot-World/","title":"AgiBot World Colosseo official platform description","publisher":"OpenDriveLab","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s5","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/","title":"Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International","publisher":"Creative Commons","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/agibot-world-alpha"},{"slug":"1x-world-model-tokenized","name":"1X World Model Compression Challenge Dataset","publisher":{"name":"1X Technologies","company":"1x-technologies"},"released":null,"summary":"Real-world humanoid interaction recordings distributed as tokenized visual observations and robot-state arrays for the 1X World Model Compression Challenge. The current v2.0 training and validation release uses NVIDIA Cosmos tokens; older v1.1 data used MAGVIT2. The tokenized dataset is Apache-2.0, while the separate raw-video release is non-commercial.","license":"Apache-2.0","commercialUse":true,"size":{"episodes":null,"hours":null,"trajectories":null,"storageTB":null,"note":"Current card publishes ~100 hours for train/val v2.0, 450 test samples and 27.3 GB total repository files. No exact episode total, duration or TB conversion is stored. The original challenge repository describes EVE, but the v2.0 card does not identify the current robot model; no catalog robot slug is assigned."},"robots":[{"robot":null,"name":"1X humanoid (current v2.0 model unspecified)"}],"modalities":["rgb","proprioception"],"format":["Binary arrays","JSON"],"realOrSim":"real","humanoid":true,"links":{"homepage":"https://huggingface.co/datasets/1x-technologies/world_model_tokenized_data","download":"https://huggingface.co/datasets/1x-technologies/world_model_tokenized_data","paper":null},"sources":[{"id":"s1","url":"https://huggingface.co/datasets/1x-technologies/world_model_tokenized_data","title":"1X World Model Compression Challenge tokenized dataset card","publisher":"1X Technologies","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://github.com/1x-technologies/1xgpt","title":"1X World Model Challenge official repository","publisher":"1X Technologies","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s3","url":"https://www.apache.org/licenses/LICENSE-2.0","title":"Apache License, Version 2.0","publisher":"Apache Software Foundation","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/1x-world-model-tokenized"},{"slug":"agibot-world-2026","name":"AgiBot World 2026","publisher":{"name":"AgiBot","company":"agibot"},"released":null,"summary":"AgiBot's 2026 dataset collected in 100% real-world environments on the AGIBOT G2 platform, with simulation data released separately in the GenieSim project. The CC BY-NC-SA 4.0 license bars commercial use.","license":"CC BY-NC-SA 4.0","commercialUse":false,"size":{"episodes":null,"hours":null,"trajectories":null,"storageTB":13.6,"note":"Hugging Face lists 13.6 TB total repository files; the card offers a ~7 GB sample. No complete episode count or release date is stated. Simulation data is distributed separately through GenieSim."},"robots":[{"robot":"agibot-g2","name":"AGIBOT G2"}],"modalities":["rgb","depth","proprioception","language"],"format":["LeRobot","Parquet","MP4"],"realOrSim":"real","humanoid":true,"links":{"homepage":"https://huggingface.co/datasets/agibot-world/AgiBotWorld2026","download":"https://huggingface.co/datasets/agibot-world/AgiBotWorld2026","paper":null},"sources":[{"id":"s1","url":"https://huggingface.co/datasets/agibot-world/AgiBotWorld2026","title":"AgiBotWorld2026 dataset card","publisher":"AgiBot","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/","title":"Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International","publisher":"Creative Commons","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/agibot-world-2026"},{"slug":"nvidia-arena-gr1-manipulation","name":"Arena-GR1-Manipulation-Task","publisher":{"name":"NVIDIA Corporation","company":null},"released":null,"summary":"Isaac Lab simulation dataset for opening a microwave with a Fourier GR1 humanoid. NVIDIA distributes human-annotated seed demonstrations, MimicGen-generated trajectories and a GR00T LeRobot conversion, with visual, state, language and action information.","license":"CC-BY-4.0","commercialUse":true,"size":{"episodes":null,"hours":null,"trajectories":50,"storageTB":null,"note":"Published 50-trajectory figure refers to the MimicGen-generated dataset; a separate annotated HDF5 file contains 10 seed demonstrations. Storage is published as 5.16 GB; no TB conversion or sum of seeds and generated data is stored. 2025-10-10 is the creation date, not a verified release date."},"robots":[{"robot":"fourier-gr-1","name":"Fourier GR1 (Isaac Lab simulation)"}],"modalities":["rgb","proprioception","language"],"format":["HDF5","LeRobot"],"realOrSim":"sim","humanoid":true,"links":{"homepage":"https://huggingface.co/datasets/nvidia/Arena-GR1-Manipulation-Task","download":"https://huggingface.co/datasets/nvidia/Arena-GR1-Manipulation-Task","paper":null},"sources":[{"id":"s1","url":"https://huggingface.co/datasets/nvidia/Arena-GR1-Manipulation-Task","title":"Arena-GR1-Manipulation-Task dataset card","publisher":"NVIDIA Corporation","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://creativecommons.org/licenses/by/4.0/","title":"Creative Commons Attribution 4.0 International license","publisher":"Creative Commons","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/nvidia-arena-gr1-manipulation"},{"slug":"ario","name":"ARIO (All Robots In One)","publisher":{"name":"ARIO Alliance","company":null},"released":null,"summary":"ARIO combines real-world robot recordings, converted datasets and simulated trajectories under a common data standard. Published examples include AgileX Cobot Magic manipulation, simulated navigation and a small share of humanoid data. Licensing depends on the constituent data source.","license":null,"commercialUse":null,"size":{"episodes":null,"hours":null,"trajectories":null,"storageTB":null,"note":"Project publishes approximately 3 million episodes from 258 series and 321,064 tasks, not an exact episode total. Team-collected data uses CC BY 4.0 or MIT; converted third-party data retains original licenses, so no single catalog-wide license is assigned."},"robots":[{"robot":null,"name":"AgileX Cobot Magic"}],"modalities":[],"format":[],"realOrSim":"mixed","humanoid":true,"links":{"homepage":"https://imaei.github.io/project_pages/ario/","download":null,"paper":"https://arxiv.org/abs/2408.10899"},"sources":[{"id":"s1","url":"https://imaei.github.io/project_pages/ario/","title":"ARIO project page (dataset, license)","publisher":"ARIO authors","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://arxiv.org/abs/2408.10899","title":"All Robots in One: A New Standard and Unified Dataset for Versatile, General-Purpose Embodied Agents","publisher":"arXiv","date":"2024-08-20","type":"paper","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/ario"},{"slug":"behavior-1k-2025-challenge-demos","name":"BEHAVIOR-1K 2025 Challenge Demos","publisher":{"name":"BEHAVIOR-1K team (Stanford)","company":"stanford-university"},"released":null,"summary":"Simulated household demonstrations from the BEHAVIOR-1K 2025 challenge: 10,000 episodes recorded with the R1Pro embodiment, 9 cameras at 30 fps, in LeRobot v2.1 format. The dataset card is tagged MIT.","license":"MIT","commercialUse":true,"size":{"episodes":10000,"hours":null,"trajectories":null,"storageTB":1.95,"note":"10,000 episodes, 119,094,660 frames and 90,000 video files across 50 tasks; 30 fps. Wrist videos are 480x480 and other camera views are 720x720; the preview header only reports 720x720. Hugging Face lists 1.95 TB total repository files. This is the 2025 simulated R1 Pro configuration. Organizer advertises 1,200+ hours for the broader demonstration release; that lower bound is not stored as an exact duration for this conversion."},"robots":[{"robot":"galaxea-r1","name":"Galaxea R1 Pro (2025 Challenge simulation)"}],"modalities":["rgb","proprioception"],"format":["LeRobot","Parquet","MP4"],"realOrSim":"sim","humanoid":true,"links":{"homepage":"https://behavior.stanford.edu/","download":"https://huggingface.co/datasets/behavior-1k/2025-challenge-demos","paper":"https://arxiv.org/abs/2403.09227"},"sources":[{"id":"s1","url":"https://behavior.stanford.edu/","title":"BEHAVIOR-1K project page","publisher":"Stanford","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://huggingface.co/datasets/behavior-1k/2025-challenge-demos","title":"behavior-1k/2025-challenge-demos dataset card","publisher":"BEHAVIOR-1K","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s3","url":"https://arxiv.org/abs/2403.09227","title":"BEHAVIOR-1K: A Human-Centered, Embodied AI Benchmark with 1,000 Everyday Activities and Realistic Simulation","publisher":"arXiv","date":"2024-03-14","type":"paper","accessed":"2026-10-13"},{"id":"s4","url":"https://behavior.stanford.edu/challenge/archive/2025/index.html","title":"2025 BEHAVIOR Challenge official dataset description","publisher":"Stanford Vision and Learning Lab","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s5","url":"https://www.linkedin.com/posts/galaxeadynamics_launched-by-fei-fei-lis-team-at-stanford-activity-7369261827798532098-sks5","title":"Galaxea official announcement: R1 and R1 Pro in the 2025 BEHAVIOR Challenge","publisher":"Galaxea","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s6","url":"https://opensource.org/license/mit","title":"The MIT License","publisher":"Open Source Initiative","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/behavior-1k-2025-challenge-demos"},{"slug":"behavior-1k-2026-challenge-demos","name":"BEHAVIOR-1K 2026 Challenge Demos","publisher":{"name":"BEHAVIOR-1K Contributors","company":"stanford-university"},"released":null,"summary":"Simulated household demonstration trajectories for the 2026 BEHAVIOR Challenge, recorded with the R1Pro embodiment and distributed in LeRobot v3 format. The release covers 100 tasks with RGB, depth, robot-state and language annotations. The dataset card and included license file specify MIT.","license":"MIT","commercialUse":true,"size":{"episodes":20000,"hours":null,"trajectories":null,"storageTB":3.26,"note":"Card and metadata publish 20,000 episodes, 210,916,774 frames, 100 tasks, 955 Parquet shards and 17,093 MP4 files. Card overview says approximately 3.0 TB; Hugging Face lists 3.26 TB total repository files. The model is identified as R1Pro in meta/info.json. First data-release date is not established."},"robots":[{"robot":"galaxea-r1","name":"Galaxea R1 Pro (OmniGibson challenge simulation)"}],"modalities":["rgb","depth","proprioception","language"],"format":["LeRobot","Parquet","MP4"],"realOrSim":"sim","humanoid":true,"links":{"homepage":"https://huggingface.co/datasets/behavior-1k/2026-challenge-demos","download":"https://huggingface.co/datasets/behavior-1k/2026-challenge-demos","paper":null},"sources":[{"id":"s1","url":"https://huggingface.co/datasets/behavior-1k/2026-challenge-demos","title":"BEHAVIOR-1K 2026 Challenge Demos official dataset card","publisher":"BEHAVIOR-1K Contributors","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://huggingface.co/datasets/behavior-1k/2026-challenge-demos/blob/main/LICENSE","title":"BEHAVIOR-1K 2026 dataset MIT license","publisher":"BEHAVIOR-1K Contributors","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s3","url":"https://huggingface.co/datasets/behavior-1k/2026-challenge-demos/blob/main/meta/info.json","title":"BEHAVIOR-1K 2026 dataset schema and statistics","publisher":"BEHAVIOR-1K Contributors","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s4","url":"https://behavior.stanford.edu/challenge/evaluation.html","title":"2026 BEHAVIOR Challenge: default R1Pro simulation embodiment","publisher":"Stanford Vision and Learning Lab","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s6","url":"https://www.linkedin.com/posts/galaxeadynamics_launched-by-fei-fei-lis-team-at-stanford-activity-7369261827798532098-sks5","title":"Galaxea official identification of the BEHAVIOR R1 Pro platform","publisher":"Galaxea","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s7","url":"https://opensource.org/license/mit","title":"The MIT License","publisher":"Open Source Initiative","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/behavior-1k-2026-challenge-demos"},{"slug":"bridgedata-v2","name":"BridgeData V2","publisher":{"name":"BridgeData V2 authors","company":null},"released":null,"summary":"Real-robot manipulation dataset of WidowX 250 demonstrations across 24 environments and 13 skills, mixing teleoperated demonstrations with scripted pick-and-place rollouts. The data is provided under CC BY 4.0.","license":"CC BY 4.0","commercialUse":true,"size":{"episodes":null,"hours":null,"trajectories":60096,"storageTB":null,"note":"50,365 teleoperated demonstrations and 9,731 scripted-policy rollouts, totaling the published 60,096 trajectories. Images are stored as JPEG; depth is present for a subset of recordings."},"robots":[{"robot":null,"name":"WidowX 250"}],"modalities":["rgb","depth","language"],"format":["JPEG"],"realOrSim":"real","humanoid":false,"links":{"homepage":"https://rail-berkeley.github.io/bridgedata/","download":"https://rail.eecs.berkeley.edu/datasets/bridge_release/data/","paper":"https://arxiv.org/abs/2308.12952"},"sources":[{"id":"s1","url":"https://rail-berkeley.github.io/bridgedata/","title":"BridgeData V2 project page","publisher":"BridgeData V2 authors","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://arxiv.org/abs/2308.12952","title":"BridgeData V2: A Dataset for Robot Learning at Scale","publisher":"arXiv","date":"2023-08-24","type":"paper","accessed":"2026-10-13"},{"id":"s3","url":"https://rail.eecs.berkeley.edu/datasets/bridge_release/data/","title":"BridgeData V2 official download directory","publisher":"UC Berkeley RAIL","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s4","url":"https://creativecommons.org/licenses/by/4.0/","title":"Creative Commons Attribution 4.0 International","publisher":"Creative Commons","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/bridgedata-v2"},{"slug":"droid","name":"DROID","publisher":{"name":"DROID project","company":null},"released":null,"summary":"DROID is a real-world manipulation dataset collected with Franka Panda arms across 564 scenes. The project reports 76k demonstration trajectories and 350 hours, and distributes RLDS training data and higher-resolution raw recordings. A dataset-wide license was not found in the opened official documentation.","license":null,"commercialUse":null,"size":{"episodes":null,"hours":350,"trajectories":76000,"storageTB":1.7,"note":"Project publishes 76k trajectories and 350 hours. Official docs list 1.7 TB for full RLDS data, 8.7 TB for stereo-HD raw recordings, and 5.6 TB when excluding stereo videos and SVO files. These are alternative representations, not additive sizes. Raw data includes stereo/depth information."},"robots":[{"robot":null,"name":"Franka Panda"}],"modalities":["rgb","proprioception","language"],"format":["RLDS","TFDS","MP4"],"realOrSim":"real","humanoid":false,"links":{"homepage":"https://droid-dataset.github.io/","download":"https://droid-dataset.github.io/droid/the-droid-dataset.html","paper":"https://arxiv.org/abs/2403.12945"},"sources":[{"id":"s1","url":"https://droid-dataset.github.io/","title":"DROID project page","publisher":"DROID project","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://github.com/droid-dataset/droid","title":"DROID robot platform repository","publisher":"DROID project","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s3","url":"https://arxiv.org/abs/2403.12945","title":"DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset","publisher":"arXiv","date":"2024-03-19","type":"paper","accessed":"2026-10-13"},{"id":"s4","url":"https://droid-dataset.github.io/droid/the-droid-dataset.html","title":"The DROID Dataset: download formats and RLDS schema","publisher":"DROID project","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/droid"},{"slug":"droid-1-0-1-lerobot","name":"DROID 1.0.1 (LeRobot conversion)","publisher":{"name":"cadene","company":null},"released":null,"summary":"Community LeRobot v2.1 conversion published by cadene under the name droid_1.0.1, with Franka robot state, three camera streams and language-instruction fields. Its card is tagged Apache-2.0, but upstream rights and the relationship between its episode count and the original DROID release remain unconfirmed.","license":"Apache-2.0","commercialUse":null,"size":{"episodes":95600,"hours":null,"trajectories":null,"storageTB":null,"note":"Conversion card publishes 95,600 episodes and 27,612,581 frames, with 399 GB total repository files. No TB conversion is stored. This differs from the upstream project’s published 76k demonstrations; the card does not explain the relationship or confirm upstream licensing."},"robots":[{"robot":null,"name":"Franka Panda"}],"modalities":["rgb","proprioception","language"],"format":["LeRobot","Parquet","MP4"],"realOrSim":"real","humanoid":false,"links":{"homepage":"https://huggingface.co/datasets/cadene/droid_1.0.1","download":"https://huggingface.co/datasets/cadene/droid_1.0.1","paper":null},"sources":[{"id":"s1","url":"https://huggingface.co/datasets/cadene/droid_1.0.1","title":"cadene/droid_1.0.1 dataset card","publisher":"Hugging Face","date":null,"type":"other","accessed":"2026-10-13"},{"id":"s2","url":"https://droid-dataset.github.io/","title":"DROID project page (official upstream reference)","publisher":"DROID project","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s3","url":"https://www.apache.org/licenses/LICENSE-2.0","title":"Apache License, Version 2.0","publisher":"Apache Software Foundation","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/droid-1-0-1-lerobot"},{"slug":"egodex","name":"EgoDex","publisher":{"name":"Apple","company":null},"released":null,"summary":"Egocentric human manipulation recordings collected with Apple Vision Pro, combining video, tracked head/body/hand poses and language annotations across 194 tabletop tasks. This is human demonstration data rather than humanoid-robot recordings, intended for learning and evaluating dexterous manipulation.","license":"CC-by-NC-ND","commercialUse":false,"size":{"episodes":null,"hours":829,"trajectories":null,"storageTB":null,"note":"Official README and current paper publish 829 hours at 30 Hz across 194 tasks. Download parts are individually listed as 300 GB, 16 GB and 200 GB; no summed storage or TB conversion is stored. README states CC-by-NC-ND without a version. Its LICENSE file governs software only, so no dataset-license version is inferred."},"robots":[],"modalities":["rgb","human-pose","language"],"format":["HDF5","MP4"],"realOrSim":"real","humanoid":false,"links":{"homepage":"https://github.com/apple-aiml-research/ml-egodex","download":"https://github.com/apple-aiml-research/ml-egodex","paper":"https://arxiv.org/abs/2505.11709"},"sources":[{"id":"s1","url":"https://github.com/apple-aiml-research/ml-egodex","title":"EgoDex official dataset README: formats, downloads and dataset terms","publisher":"Apple","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://github.com/apple-aiml-research/ml-egodex/blob/main/LICENSE","title":"EgoDex software license (not the dataset license)","publisher":"Apple","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s3","url":"https://arxiv.org/abs/2505.11709","title":"EgoDex: Learning Dexterous Manipulation from Large-Scale Egocentric Video","publisher":"arXiv","date":"2025-05-16","type":"paper","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/egodex"},{"slug":"fourier-actionnet","name":"Fourier ActionNet Dataset","publisher":{"name":"Fourier","company":"fourier"},"released":null,"summary":"Fourier's humanoid manipulation dataset for GR1 and GR2 robots, stored as HDF5 robot logs with RGB and depth video per episode. The Hugging Face repository is gated, and its card license is non-commercial.","license":"CC BY-NC-SA 4.0","commercialUse":false,"size":{"episodes":null,"hours":null,"trajectories":null,"storageTB":2.74,"note":"Hugging Face lists 2.74 TB total repository files and a 10K-100K size category, not an exact episode count. Raw logs and videos can be converted to LeRobot v2. Access is gated."},"robots":[{"robot":"fourier-gr-2","name":"GR2"},{"robot":null,"name":"GR1-T1 / GR1-T2"}],"modalities":["rgb","depth","proprioception","language"],"format":["HDF5","MP4","MKV","JSON"],"realOrSim":"real","humanoid":true,"links":{"homepage":"https://huggingface.co/datasets/FourierIntelligence/ActionNet","download":"https://huggingface.co/datasets/FourierIntelligence/ActionNet","paper":null},"sources":[{"id":"s1","url":"https://huggingface.co/datasets/FourierIntelligence/ActionNet","title":"FourierIntelligence/ActionNet dataset card","publisher":"Fourier","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/","title":"Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International","publisher":"Creative Commons","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/fourier-actionnet"},{"slug":"unitree-g1-wbt-clothes-basket","name":"G1_WBT_Inspire_Put_Clothes_Into_Basket","publisher":{"name":"Unitree Robotics","company":"unitree"},"released":null,"summary":"Unitree G1 whole-body teleoperation recordings for placing clothes into a basket, collected with Inspire hands as part of the UnifoLM-WBT collection. The release contains synchronized camera streams, body and hand states, and actions in LeRobot v3.0 format. Unitree’s official history identifies this collection as real-robot data.","license":"Apache-2.0","commercialUse":true,"size":{"episodes":300,"hours":null,"trajectories":null,"storageTB":null,"note":"Card publishes 300 episodes, 228,479 rows and 12.4 GB total repository files; no duration or TB conversion is stored. Unitree dates the collection launch to 2026-03-05, but this task subset’s initial release date is unspecified."},"robots":[{"robot":"unitree-g1","name":"Unitree G1 with Inspire hands"}],"modalities":["rgb","proprioception"],"format":["LeRobot","Parquet","MP4"],"realOrSim":"real","humanoid":true,"links":{"homepage":"https://huggingface.co/datasets/unitreerobotics/G1_WBT_Inspire_Put_Clothes_Into_Basket","download":"https://huggingface.co/datasets/unitreerobotics/G1_WBT_Inspire_Put_Clothes_Into_Basket","paper":null},"sources":[{"id":"s1","url":"https://huggingface.co/datasets/unitreerobotics/G1_WBT_Inspire_Put_Clothes_Into_Basket","title":"G1_WBT_Inspire_Put_Clothes_Into_Basket dataset card","publisher":"Unitree Robotics","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://huggingface.co/collections/unitreerobotics/unifolm-wbt-dataset","title":"UnifoLM_WBT_Dataset official collection membership","publisher":"Unitree Robotics","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s3","url":"https://www.unitree.com/cn/about/","title":"Unitree history: real-world UnifoLM-WBT data release","publisher":"Unitree Robotics","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s4","url":"https://www.apache.org/licenses/LICENSE-2.0","title":"Apache License, Version 2.0","publisher":"Apache Software Foundation","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/unitree-g1-wbt-clothes-basket"},{"slug":"hiw-500","name":"HIW-500 (LeRobot)","publisher":{"name":"BitRobot","company":null},"released":null,"summary":"LeRobot v3.0 conversion of BitRobot’s HIW-500 whole-body teleoperation recordings on Unitree G1 robots in real homes in Southeast Asia. It includes camera observations, robot state, IMU and language annotations. The public release is licensed CC BY 4.0.","license":"CC BY 4.0","commercialUse":true,"size":{"episodes":23743,"hours":null,"trajectories":null,"storageTB":2.15,"note":"LeRobot metadata publishes 23,743 episodes and 40,839,947 frames; card overview says 500+ hours and around 10 TB. Hugging Face lists 2.15 TB for this compressed conversion. No release date is stated; the citation year is 2026."},"robots":[{"robot":"unitree-g1","name":"G1"}],"modalities":["rgb","proprioception","imu","language"],"format":["LeRobot","Parquet","MP4"],"realOrSim":"real","humanoid":true,"links":{"homepage":"https://huggingface.co/datasets/BitRobot/HIW-500-LeRobot","download":"https://huggingface.co/datasets/BitRobot/HIW-500-LeRobot","paper":null},"sources":[{"id":"s1","url":"https://huggingface.co/datasets/BitRobot/HIW-500-LeRobot","title":"BitRobot/HIW-500-LeRobot dataset card","publisher":"BitRobot","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://creativecommons.org/licenses/by/4.0/","title":"Creative Commons Attribution 4.0 International","publisher":"Creative Commons","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/hiw-500"},{"slug":"humanoid-everyday","name":"Humanoid Everyday","publisher":{"name":"University of Southern California / Toyota Research Institute","company":null},"released":null,"summary":"Open-world humanoid manipulation dataset of about 10,300 trajectories and over 3 million frames across 260 tasks, recorded on Unitree G1 and H1 with RGB-D, LiDAR, tactile and language annotations. The project page shows no download link and no dataset license.","license":null,"commercialUse":null,"size":{"episodes":null,"hours":null,"trajectories":10300,"storageTB":null,"note":"Project and paper publish 10.3k trajectories, over 3 million frames and 260 tasks across seven categories. The project page provides no dataset download or license, despite the paper stating that they are publicly available; its evaluation link is still marked Coming Soon. Paper publication is not a verified data-release date."},"robots":[{"robot":"unitree-g1","name":"G1 (Dex3-1 hands)"},{"robot":"unitree-h1","name":"H1 (INSPIRE hands)"}],"modalities":["rgb","depth","lidar","tactile","imu","proprioception","language"],"format":[],"realOrSim":"real","humanoid":true,"links":{"homepage":"https://humanoideveryday.github.io","download":null,"paper":"https://arxiv.org/abs/2510.08807"},"sources":[{"id":"s1","url":"https://humanoideveryday.github.io","title":"Humanoid Everyday project page","publisher":"Humanoid Everyday authors","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://arxiv.org/abs/2510.08807","title":"Humanoid Everyday: A Comprehensive Robotic Dataset for Open-World Humanoid Manipulation","publisher":"arXiv","date":"2025-10-09","type":"paper","accessed":"2026-10-13"},{"id":"s3","url":"https://arxiv.org/html/2510.08807","title":"Humanoid Everyday paper: platforms and multimodal collection","publisher":"arXiv","date":"2026-07-04","type":"paper","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/humanoid-everyday"},{"slug":"lerobot-aloha-static-coffee","name":"LeRobot ALOHA Static Coffee","publisher":{"name":"LeRobot (Hugging Face)","company":"hugging-face"},"released":null,"summary":"LeRobot-hosted set of 50 ALOHA demonstrations (about 55,000 frames) for a coffee task, recorded on bimanual ALOHA arms with four cameras. The card is tagged MIT, so commercial use is allowed.","license":"MIT","commercialUse":true,"size":{"episodes":50,"hours":null,"trajectories":null,"storageTB":null,"note":"Card publishes 50 episodes, 55,000 frames, four cameras and 1.57 GB total repository files. No TB conversion is stored. The card is tagged MIT; the cited ALOHA paper establishes real-robot bimanual demonstrations."},"robots":[{"robot":null,"name":"ALOHA (bimanual)"}],"modalities":["rgb","proprioception"],"format":["LeRobot","Parquet","MP4"],"realOrSim":"real","humanoid":false,"links":{"homepage":"https://huggingface.co/datasets/lerobot/aloha_static_coffee","download":"https://huggingface.co/datasets/lerobot/aloha_static_coffee","paper":"https://arxiv.org/abs/2304.13705"},"sources":[{"id":"s1","url":"https://huggingface.co/datasets/lerobot/aloha_static_coffee","title":"lerobot/aloha_static_coffee dataset card","publisher":"Hugging Face (LeRobot)","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://arxiv.org/abs/2304.13705","title":"Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware","publisher":"arXiv","date":"2023-04-23","type":"paper","accessed":"2026-10-13"},{"id":"s3","url":"https://opensource.org/license/mit","title":"The MIT License","publisher":"Open Source Initiative","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/lerobot-aloha-static-coffee"},{"slug":"libero-demos","name":"LIBERO Demonstrations","publisher":{"name":"LIBERO project","company":null},"released":null,"summary":"Human teleoperation demonstrations collected in simulation for LIBERO’s four manipulation task suites. Each task has 50 demonstration trajectories for a simulated Franka Panda arm, with visual observations, robot state and task language. The project explicitly licenses the datasets under CC BY 4.0.","license":"CC BY 4.0","commercialUse":true,"size":{"episodes":null,"hours":null,"trajectories":null,"storageTB":null,"note":"Paper publishes 50 demonstration trajectories per task across 130 tasks; no computed aggregate is stored. Official mirror lists 100 GB, not TB. Its Apache-2.0 card tag conflicts with the project README’s explicit CC BY 4.0 dataset license; the explicit project license is used. MIT applies to code only."},"robots":[{"robot":null,"name":"Franka Panda (simulated)"}],"modalities":["rgb","proprioception","language"],"format":["HDF5"],"realOrSim":"sim","humanoid":false,"links":{"homepage":"https://github.com/Lifelong-Robot-Learning/LIBERO","download":"https://huggingface.co/datasets/yifengzhu-hf/LIBERO-datasets","paper":"https://arxiv.org/abs/2306.03310"},"sources":[{"id":"s1","url":"https://github.com/Lifelong-Robot-Learning/LIBERO","title":"LIBERO repository README (license, dataset download)","publisher":"LIBERO project","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://arxiv.org/abs/2306.03310","title":"LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning","publisher":"arXiv","date":"2023-06-05","type":"paper","accessed":"2026-10-13"},{"id":"s3","url":"https://arxiv.org/html/2306.03310","title":"LIBERO paper, HTML version (describes the simulation benchmark)","publisher":"arXiv","date":null,"type":"paper","accessed":"2026-10-13"},{"id":"s4","url":"https://huggingface.co/datasets/yifengzhu-hf/LIBERO-datasets","title":"Official LIBERO HDF5 dataset mirror","publisher":"LIBERO authors","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s6","url":"https://github.com/Lifelong-Robot-Learning/LIBERO/blob/master/libero/libero/envs/bddl_base_domain.py","title":"LIBERO environment: robot-state sensor observations","publisher":"LIBERO project","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s7","url":"https://raw.githubusercontent.com/Lifelong-Robot-Learning/LIBERO/master/libero/libero/envs/env_wrapper.py","title":"LIBERO simulation wrapper: default Panda robot and RGB cameras","publisher":"LIBERO project","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s8","url":"https://creativecommons.org/licenses/by/4.0/","title":"Creative Commons Attribution 4.0 International","publisher":"Creative Commons","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/libero-demos"},{"slug":"nvidia-gr00t-teleop-sim","name":"NVIDIA GR00T Teleop Sim (GR1)","publisher":{"name":"NVIDIA GEAR","company":null},"released":null,"summary":"NVIDIA's simulated GR1 humanoid teleoperation set: 1,000 trajectories for each of 24 tabletop tasks, with upper-body control and simulated Fourier hands. The card license is CC BY-NC 4.0 and the card states the set is ready for non-commercial use.","license":"CC BY-NC 4.0","commercialUse":false,"size":{"episodes":null,"hours":null,"trajectories":null,"storageTB":null,"note":"Card publishes 1,000 trajectories for each of 24 tabletop tasks; no computed total is stored. Its opening paragraph separately says 1,000 trajectories. Published storage is ~14 GB HDF5, ~39 GB LeRobot-style, and 55.4 GB total repository files; no TB conversion is stored. June 1, 2025 is the creation date, not a confirmed release date."},"robots":[{"robot":"fourier-gr-1","name":"GR1 (simulated Fourier hands)"}],"modalities":["rgb","proprioception","language"],"format":["HDF5","LeRobot"],"realOrSim":"sim","humanoid":true,"links":{"homepage":"https://huggingface.co/datasets/nvidia/PhysicalAI-Robotics-GR00T-Teleop-Sim","download":"https://huggingface.co/datasets/nvidia/PhysicalAI-Robotics-GR00T-Teleop-Sim","paper":null},"sources":[{"id":"s1","url":"https://huggingface.co/datasets/nvidia/PhysicalAI-Robotics-GR00T-Teleop-Sim","title":"PhysicalAI-Robotics-GR00T-Teleop-Sim dataset card","publisher":"NVIDIA","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://creativecommons.org/licenses/by-nc/4.0/","title":"Creative Commons Attribution-NonCommercial 4.0 International","publisher":"Creative Commons","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/nvidia-gr00t-teleop-sim"},{"slug":"open-x-embodiment","name":"Open X-Embodiment","publisher":{"name":"Open X-Embodiment Collaboration","company":null},"released":null,"summary":"Open X-Embodiment pools real-robot manipulation datasets from multiple institutions, originally spanning 22 robot embodiments. It contains more than one million trajectories in standardized robot-learning formats. No single license or commercial-use permission is asserted here for all constituent datasets.","license":null,"commercialUse":null,"size":{"episodes":null,"hours":null,"trajectories":null,"storageTB":null,"note":"Project publishes 1M+ real robot trajectories, not an exact count. Constituent datasets retain their own metadata and citations. The code repository’s Apache-2.0/CC BY 4.0 notice is not treated as a blanket dataset license."},"robots":[],"modalities":[],"format":["TFDS"],"realOrSim":"real","humanoid":false,"links":{"homepage":"https://robotics-transformer-x.github.io/","download":"https://github.com/google-deepmind/open_x_embodiment","paper":"https://arxiv.org/abs/2310.08864"},"sources":[{"id":"s1","url":"https://robotics-transformer-x.github.io/","title":"Open X-Embodiment project page","publisher":"Open X-Embodiment Collaboration","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://github.com/google-deepmind/open_x_embodiment","title":"open_x_embodiment repository README","publisher":"Google DeepMind","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s3","url":"https://arxiv.org/abs/2310.08864","title":"Open X-Embodiment: Robotic Learning Datasets and RT-X Models","publisher":"arXiv","date":"2023-10-13","type":"paper","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/open-x-embodiment"},{"slug":"rh20t","name":"RH20T","publisher":{"name":"RH20T authors","company":null},"released":null,"summary":"Real-world manipulation dataset covering 147 tasks and over 110,000 sequences, with synchronized visual, robot-state, force-torque, audio and selected fingertip tactile data. RH20T-C is licensed CC BY-SA 4.0. RH20T-NC prohibits commercial use, but its stated CC BY-NC 4.0 label links to CC BY-NC-SA 4.0; the catalog leaves the combined license and commercial-use flag unresolved.","license":null,"commercialUse":null,"size":{"episodes":null,"hours":null,"trajectories":null,"storageTB":40,"note":"Project publishes 40 TB for the original dataset, ~5 TB for resized RGB and ~10 TB for resized RGBD. The sequence count is only stated as over 110,000. Tactile sensing is limited to configuration 7. Licence conflict: RH20T-NC text names CC BY-NC 4.0, but its licence link opens CC BY-NC-SA 4.0; RH20T-C names CC BY-SA 4.0."},"robots":[],"modalities":["rgb","depth","infrared","proprioception","force-torque","audio","tactile"],"format":["TAR.GZ"],"realOrSim":"real","humanoid":false,"links":{"homepage":"https://rh20t.github.io/","download":"https://rh20t.github.io/#download","paper":"https://arxiv.org/abs/2307.00595"},"sources":[{"id":"s1","url":"https://rh20t.github.io/","title":"RH20T project page (license and size sections)","publisher":"RH20T authors","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://arxiv.org/abs/2307.00595","title":"RH20T: A Comprehensive Robotic Dataset for Learning Diverse Skills in One-Shot","publisher":"arXiv","date":"2023-07-02","type":"paper","accessed":"2026-10-13"},{"id":"s3","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/","title":"License destination linked by the RH20T-NC label","publisher":"Creative Commons","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/rh20t"},{"slug":"robomind","name":"RoboMIND","publisher":{"name":"RoboMIND Team (X-Humanoid, Peking University and Beijing Academy of Artificial Intelligence)","company":"x-humanoid"},"released":null,"summary":"RoboMIND’s current Hugging Face release includes teleoperated robot demonstrations and digital-twin simulation data, stored as HDF5 trajectories with visual observations, robot states and language instructions. The project reports 107k real-world demonstrations for versions 1.1/1.2; the initial version 1.0 was smaller. The official dataset card is tagged Apache-2.0.","license":"Apache-2.0","commercialUse":true,"size":{"episodes":null,"hours":null,"trajectories":107000,"storageTB":12.3,"note":"Published 107k figure describes real-world demonstrations in versions 1.1/1.2, not the total including simulation. Version 1.0 had 55K trajectories and 279 tasks. Version 1.2 adds one real and nine simulated Upright_Cup tasks. Hugging Face lists 12.3 TB total repository files. Release dates for these versions are not stated."},"robots":[{"robot":null,"name":"Franka Emika Panda"},{"robot":null,"name":"UR5e"},{"robot":null,"name":"AgileX Cobot Magic V2.0"},{"robot":null,"name":"Tien Kung humanoid (model not stated)"}],"modalities":["rgb","proprioception","language"],"format":["HDF5"],"realOrSim":"mixed","humanoid":true,"links":{"homepage":"https://x-humanoid-robomind.github.io/","download":"https://huggingface.co/datasets/x-humanoid-robomind/RoboMIND","paper":"https://arxiv.org/abs/2412.13877"},"sources":[{"id":"s1","url":"https://x-humanoid-robomind.github.io/","title":"RoboMIND project page","publisher":"X-Humanoid","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://arxiv.org/abs/2412.13877","title":"RoboMIND: Benchmark on Multi-embodiment Intelligence Normative Data for Robot Manipulation","publisher":"arXiv","date":"2024-12-18","type":"paper","accessed":"2026-10-13"},{"id":"s3","url":"https://huggingface.co/datasets/x-humanoid-robomind/RoboMIND","title":"Official RoboMIND dataset card: license, versions and formats","publisher":"RoboMIND Team","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s4","url":"https://www.apache.org/licenses/LICENSE-2.0","title":"Apache License, Version 2.0","publisher":"Apache Software Foundation","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/robomind"},{"slug":"robomind-2-0","name":"RoboMIND 2.0","publisher":{"name":"RoboMIND Team (X-Humanoid, Peking University and Beijing Academy of Artificial Intelligence)","company":"x-humanoid"},"released":null,"summary":"RoboMIND 2.0 combines real-world bimanual and mobile-manipulation trajectories across six robot platforms with a separate simulated dataset. Its latest paper documents HDF5 files and links the public release to ModelScope. A dataset license could not be verified from the text-only ModelScope pages.","license":null,"commercialUse":null,"size":{"episodes":null,"hours":null,"trajectories":null,"storageTB":null,"note":"Latest paper publishes over 310K real trajectories, including 12K tactile-enhanced episodes and 20K mobile trajectories, plus an additional 20K simulated trajectories. Exact combined count and storage are not published. Its abstract says 739 tasks while the main text says 759. The main collection identifies Tian Yi and Tien Kung without a generation; their catalog links are left null."},"robots":[{"robot":null,"name":"Franka (dual-arm)"},{"robot":null,"name":"UR5e (dual-arm)"},{"robot":null,"name":"AgileX"},{"robot":null,"name":"ARX"},{"robot":null,"name":"Tian Yi (collection generation unspecified)"},{"robot":null,"name":"Tien Kung (collection generation unspecified)"}],"modalities":["rgb","depth","tactile","force-torque","proprioception","language"],"format":["HDF5"],"realOrSim":"mixed","humanoid":true,"links":{"homepage":"https://x-humanoid-robomind.github.io/","download":"https://www.modelscope.cn/collections/X-Humanoid/RoboMIND20","paper":"https://arxiv.org/abs/2512.24653"},"sources":[{"id":"s1","url":"https://arxiv.org/abs/2512.24653","title":"RoboMIND 2.0: A Multimodal, Bimanual Mobile Manipulation Dataset for Generalizable Embodied Intelligence","publisher":"arXiv","date":"2025-12-31","type":"paper","accessed":"2026-10-13"},{"id":"s2","url":"https://arxiv.org/html/2512.24653v3","title":"RoboMIND 2.0 paper, latest version: platforms, HDF5 format and release links","publisher":"arXiv","date":"2026-02-27","type":"paper","accessed":"2026-10-13"},{"id":"s3","url":"https://x-humanoid-robomind.github.io/","title":"RoboMIND project: official v2.0 release announcement","publisher":"RoboMIND Team","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s4","url":"https://www.modelscope.cn/collections/X-Humanoid/RoboMIND20","title":"RoboMIND 2.0 official ModelScope collection","publisher":"X-Humanoid","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/robomind-2-0"},{"slug":"robotwin-2","name":"RoboTwin 2.0 Dataset","publisher":{"name":"RoboTwin team","company":null},"released":null,"summary":"Simulation-generated bimanual manipulation demonstrations for RoboTwin 2.0, with Aloha-AgileX as the default embodiment. Official documentation describes pre-collected HDF5 trajectories, and the download repository also contains LeRobot conversions. The dataset repository is tagged MIT.","license":"MIT","commercialUse":true,"size":{"episodes":null,"hours":null,"trajectories":null,"storageTB":1.53,"note":"Hugging Face lists 1.53 TB for the whole repository, which also contains assets and checkpoints. Official collection docs state over 100,000 pre-collected trajectories, not an exact count. Native HDF5 and LeRobot versions are distributed; Parquet auto-viewer metadata was not used to identify the native format."},"robots":[{"robot":null,"name":"Aloha-AgileX (simulated default bimanual embodiment)"},{"robot":null,"name":"Franka (simulated dual-arm; model unspecified)"},{"robot":null,"name":"UR5 (simulated dual-arm)"},{"robot":null,"name":"Piper (simulated dual-arm)"},{"robot":null,"name":"ARX-X5 (simulated dual-arm)"}],"modalities":["rgb","proprioception","language"],"format":["HDF5","LeRobot"],"realOrSim":"sim","humanoid":false,"links":{"homepage":"https://robotwin-platform.github.io/","download":"https://huggingface.co/datasets/TianxingChen/RoboTwin2.0","paper":"https://arxiv.org/abs/2506.18088"},"sources":[{"id":"s1","url":"https://huggingface.co/datasets/TianxingChen/RoboTwin2.0","title":"TianxingChen/RoboTwin2.0 dataset repository","publisher":"RoboTwin authors","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s2","url":"https://arxiv.org/abs/2506.18088","title":"RoboTwin 2.0 paper (simulation-based data synthesis)","publisher":"arXiv","date":"2025-06-22","type":"paper","accessed":"2026-10-13"},{"id":"s3","url":"https://robotwin-platform.github.io/","title":"RoboTwin 2.0 official project","publisher":"RoboTwin Team","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s4","url":"https://robotwin-platform.github.io/doc/usage/collect-data.html","title":"RoboTwin 2.0 official dataset collection and format documentation","publisher":"RoboTwin Team","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s5","url":"https://huggingface.co/datasets/TianxingChen/RoboTwin2.0/tree/main","title":"RoboTwin 2.0 repository inventory: raw and LeRobot data","publisher":"RoboTwin Team","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s6","url":"https://github.com/RoboTwin-Platform/RoboTwin/blob/main/envs/_base_task.py","title":"RoboTwin collection implementation: RGB and robot joint observations","publisher":"RoboTwin Team","date":null,"type":"official","accessed":"2026-10-13"},{"id":"s7","url":"https://arxiv.org/html/2506.18088","title":"RoboTwin 2.0 paper: collection embodiments and language instructions","publisher":"arXiv","date":null,"type":"paper","accessed":"2026-10-13"},{"id":"s8","url":"https://opensource.org/license/mit","title":"The MIT License","publisher":"Open Source Initiative","date":null,"type":"official","accessed":"2026-10-13"}],"lastVerified":"2026-10-13","url":"https://humanoidversus.com/datasets/robotwin-2"}]}