The project describes a Geometric Context Transformer that combines coordinate grounding, dense geometric cues, and long-range drift correction, and reports inference at about 20 FPS at 518×378 resolution. These are the project’s claims, not independently verified performance results.
The repository’s news section says evaluation scripts for KITTI and Oxford Spires were released on 2026-05-25. Its documentation also describes keyframe and windowed inference for long sequences and showcases an offline-rendered indoor walkthrough of about 25,000 frames. https://github.com/Robbyant/lingbot-map