Friday, July 17, 2026

LingBot-Map

★16,056

Feed-forward 3D foundation model that rebuilds scenes from a video stream.

LingBot-Map

Why it matters

Most strong 3D reconstruction still leans on slow iterative optimization. LingBot-Map is a streaming foundation model that keeps geometric context in a feed-forward path, runs around 20 FPS on modest resolution, and stays stable over very long videos. Open weights and demos make this useful if you care about robotics, mapping, or real-time scene understanding without a heavy offline pipeline.

3dcomputer-visionfoundation-model
Source ↗

Via github

X