RealSense Dual RGB + NVIDIA Fast-FoundationStereo live depth demo hits ~8.5 fps

chrismatthieu · x · 2026-11-20

Developer chrismatthieu, inspired by LearnOpenCV's RealSense + AI post, hacked together a demo combining Intel RealSense's new Dual RGB feature with NVIDIA Robotics' Fast-FoundationStereo model for live stereo depth estimation. Screenshots show per-camera D405 depth maps and a fused depth output running CUDA PyTorch inference at 8.5 fps, with clean depth segmentation of people and objects. More to come, he says.

Original post →

More from Embodied

Embodied channel →