Building datasets for Physical AI requires capturing human manipulation at scale, but traditional approaches create a bottleneck. Full-scale data collection facilities are expensive and limited in deployment, while vision-only systems struggle with occlusion during complex hand movements. OpenGraph Labs, which operates large-scale data engines and a Physical AI Training Center in South Korea, needed a practical solution that robotics teams could deploy anywhere to start gathering high-quality egocentric data immediately.
OpenGraph Labs Co-founder and CEO Julia Kim demonstrated this with a streamlined setup pairing MANUS data gloves with iPhone egocentric vision. The MANUS data gloves capture precise hand actions with 25 degrees of freedom and millimeter-level accuracy using EMF tracking, while the iPhone handles visual context. Custom software manages clock synchronization between vision and hand data streams, with UDP communication enabling seamless data transfer. This accessible configuration allows researchers to collect synchronized multimodal data on a daily basis, transforming data collection from a specialized facility-only activity into something team scan do routinely in diverse environments.
The effectiveness of MANUS gloves in this setup stems from their combination of precision and practical deploy ability. Unlike marker-based systems that require controlled lab environments or vision-only approaches that fail during hand occlusion, MANUS gloves deliver consistent, drift-free finger tracking across any setting. The electromagnetic tracking technology captures every nuance of hand motion in real time without line-of-sight requirements. For robotics teams focused on teleoperation and manipulation, this means building datasets that accurately represent the complexity of dexterous human movements. The gloves' low latency and high fidelity, combined with straightforward integration via standard protocols, allow teams to start egocentric data collection immediately rather than waiting for access to specialized facilities.
OpenGraph Labs demonstrates how MANUS gloves enable robotics professionals to scale data collection beyond centralized infrastructure, making high-quality demonstration data for manipulation tasks both accessible and practical for distributed teams working on Physical AI.
