LingBot-Map: A Feed-Forward 3D Foundation Model for Streaming Scene Reconstruction

Lingbot-map: A 3D foundation model for reconstructing scenes from streaming data

I am excited to introduce LingBot-Map, a new feed-forward 3D foundation model designed for real-time scene reconstruction from streaming data. By unifying geometric context and long-range drift correction, our system achieves stable inference at 20 FPS on sequences exceeding 10,000 frames. This approach delivers state-of-the-art performance across diverse benchmarks, outperforming existing iterative optimization methods while maintaining high efficiency for both indoor and outdoor environments.

"We've built a feed-forward 3D foundation model for streaming 3D reconstruction that enables stable inference at ~20 FPS on long sequences exceeding 10,000 frames."