Volumetric Streaming over RTC

Realtime 3D capture of a person, merged from two sensors and streamed over the internet.

Role
Hardware selection and rig build, software architecture, engineering, demonstration
Areas
Telepresence, Hardware, Computer Vision
Two Intel RealSense depth cameras on black 3D-printed plates, mounted at either end of an aluminum rail on a small stand.

This exploration focused on realtime volumetric capture and streaming over the internet.

We selected the sensors, then designed, built and calibrated the test rig: two Intel RealSense D435 cameras on 3D-printed mounts and an 80/20 frame, positioned so each fills in what the other can't see. We built two spacings. The narrow rig sits on a monitor and fills occlusions tightly for laptop-sized screens. The wide rig spreads the sensors along a longer rail and fills much more, but only suits television-sized displays.

We wrote the software that merges the two volumes, correcting for alignment and color, and encoded the result in a format that can be streamed between remote devices and rebuilt on the far end over current network infrastructure.

It became the foundation for the rest of Enhanced Presence.