Work / Fig 06 · 2025
FBT / XR Full Body Tracking
Low-cost full-body tracking using vision-language models and ultra-low-latency multi-stream video. ~98% accuracy excluding heavy occlusion.
A jointed mannequin stands on a round platform between two cameras on tripods, their cables running to a tracking station. Tracker dots on the mannequin's joints light one after another as each frame is solved, and the station's screen draws the same skeleton.
Affordable motion capture for VR, AR, and XR
Engineered ultra-low-latency, high-bandwidth protocols for multiple concurrent live video streams with redundant pipelines. Architected consumer-grade full-body tracking on VLMs at ~98% accuracy, excluding edge-case heavy occlusion.
Built under Sigmoid Technologies with a crowdfunding and community push. Status: sunset. The technical thesis still stands: body tracking should not require a motion-capture studio.