LingBot-Map: A 3D Model for Scene Reconstruction from Streaming Data

LingBot-Map is a newly developed model for reconstructing 3D scenes from streaming data. The Geometric Context Transformer integrates coordinate grounding, dense geometric cues, and long-range drift correction into a cohesive framework. This model achieves high-efficiency streaming inference at 20 FPS, demonstrating superior performance on long sequences exceeding 10,000 frames.