Hub

Founded by Tim Sprecher in 2024

Collect egocentric video and physical sensor data to train robotics and embodied models

Pricing model

Paid

Rating

★ People love it

Time to value

Requires Expertise

You can use Hub to collect real-world physical and egocentric data for training embodied AI and robotics models. The service gathers multimodal datasets including first-person stereo video, metric depth, synchronized IMU, 3D hand tracking, and tactile sensor readings from thousands of commercial and residential locations worldwide. You can request custom collection runs specifying your required environment, sensor modalities, and diversity targets. Hub handles contributor coordination, data enrichment, frame-level quality checks, and delivers structured datasets formatted for robotic training frameworks directly to your cloud storage.

What Hub does

Request custom physical data collection for robotics modelsConfigure data diversity caps and environment requirementsReview sample capture frames directly in cloud storageExport synchronized sensor data into standard robotics formatsTrack human manipulation motions with 3D hand keypoint dataCollect first-person stereo video footage across commercial sitesCollects egocentric video with metric depth and synchronized IMU sensorsRecords 3D hand poses and finger tracking with twenty-one keypointsCaptures robot teleoperation trajectories from human hardware operatorsRecords tactile touch and contact pressure maps from instrumented glovesSources physical data across hundreds of verified commercial locationsDelivers custom data files formatted in MCAP or LeRobot standardsProvides automated frame-level quality assurance and review before full delivery

Pricing breakdown

Pricing model: Paid

Tutorials & Demos

Frequently asked

Amazon S3, LeRobot, MCAP, Project Aria (Hardware), Apple Vision Pro (Hardware), Custom API Endpoints, Project Aria, Apple Vision Pro, PyTorch DataLoaders, Hugging Face Hub

Want a tailored answer?