Mecka offers EgoVerse
Product
EgoVerse is an ecosystem for curating human egocentric data and learning from it for robot learning.
- Currently Offered
EgoVerse is an ecosystem for curating human egocentric data and learning from it for robot learning.
EgoVerse hosts a living dataset of human demonstrations continuously expanded by a consortium of academic and industrial partners. The dataset includes accurate camera poses, 3D head tracking, dense language annotation, and standardized tasks across diverse scenes, objects, and operators. With over 1,362 hours of human demos across nearly 80,000 episodes, 1,965 tasks, and 240 scenes, EgoVerse is built for human-to-robot transfer.
EgoVerse is validated across diverse robot embodiments including object-in-container at Georgia Tech, cup-on-saucer bimanual precision at Stanford, and bag-grocery long-horizon tasks at UC San Diego and ETH Zurich. The platform offers open-source dataset browser and GitHub access, driven by a research community advancing science of human-to-robot transfer.
A data lab that collects and synchronizes multimodal embodied intelligence data to train robotics and physical AI systems.
Robotics Data Infrastructure is XDOF's data pipeline, collection, and annotation platform for robotics foundation models.
Octo is an open-source transformer-based diffusion policy pretrained on 800k robot episodes for generalist manipulation.
OpenVLA is a 7B-parameter open-source vision-language-action model for generalist robot manipulation.
Ready-to-license expert-authored datasets for AI post-training across professional, coding, search, and multimodal tasks.
Instance verifies robot task completion from camera footage, returning grounded success evidence for every episode.
Datoric's build-to-spec data collection service for AI teams.
Datoric's world-model and agentic training-data catalog.
Expert-curated datasets and reinforcement learning environments for training and evaluating AI models.
MangoDesk's platform spins up bespoke annotation pipelines for AI evaluation.
A data lab that collects and synchronizes multimodal embodied intelligence data to train robotics and physical AI systems.
Datoric's robotics training-data catalog for physical AI teams.
Robotics Data Infrastructure is XDOF's data pipeline, collection, and annotation platform for robotics foundation models.
micro1 Robotics Data captures expert demonstrations for physical policy models.
Octo is an open-source transformer-based diffusion policy pretrained on 800k robot episodes for generalist manipulation.
OpenVLA is a 7B-parameter open-source vision-language-action model for generalist robot manipulation.
SignSplit is the signed-data platform for protecting, licensing, and contributing consented real-world human data.
Ready-to-license expert-authored datasets for AI post-training across professional, coding, search, and multimodal tasks.
Ango Hub is iMerit's human-in-the-loop data platform for AI development, combining annotation, workflow design, validation, and analytics.
Instance verifies robot task completion from camera footage, returning grounded success evidence for every episode.
Datoric's build-to-spec data collection service for AI teams.
Datoric's world-model and agentic training-data catalog.
Expert-curated datasets and reinforcement learning environments for training and evaluating AI models.
MangoDesk's platform spins up bespoke annotation pipelines for AI evaluation.