← Anchor3R project Interactive research preview

Same sequence. Shared viewpoint.

LingBot-Map and Anchor3R offline + loop on KITTI, with full-sequence GT trajectory alignment.

Drag any of the three views to rotate all · scroll / pinch to zoom · right-drag or Shift-drag to pan · shared camera, projection and metric scale · keyboard arrows to rotate, + / − to zoom, Home to reset.

Camera trajectories

LingBot-Map
Anchor3R (offline)
Ground truthGT · meters

Preparing full-frame trajectories…

LingBot-Map

Preparing cloud…

Anchor3R offline

Preparing cloud…

Controls

Loading comparison metadata…

Camera trajectories · static export

Download SVG

Three trajectories, all frames, one global alignment per prediction.

Full-sequence LingBot-Map, Anchor3R and GT camera trajectories, with a top-down plot and height versus frame plot.

Global point-cloud overview

Download PNG

Static fallback: 70k sampled points per method, the same bounds and camera, and a fixed 20 m cutoff. The interactive controls do not change this image.

Side-by-side global point clouds for LingBot-Map and Anchor3R using the same viewpoint and GT coordinate system.

Data & browser budget

MethodFull pointsWithin 20 mFull PLYPreview / detail

Only the selected sequence and quality are fetched. Switching releases old GPU buffers; rendering happens on interaction rather than continuously. Raw PLY and full-resolution octree archives are intentionally not bundled into this project page.

Download alignment, sampling and source-checksum metadata