3D Reconstruction on ScanNet++ (25 scenes)
0.0688Chamfer Distance (m)GenRecon
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GenRecon2026.05 | 0.0688 | 77.71 | 78.6 | |
| FineReconProcess=3D refinement of fused monocular predictions, Foundation Model=DA32026.05 | 0.0819 | 76.94 | 78.51 | |
| MurrePrior=diffusion-based depth priors, Conditioning=3D conditioning2026.05 | 0.0978 | 72.63 | 71.25 | |
| DA3Model Type=feed-forward monocular depth foundation model, Fusion=TSDF fusion2026.05 | 0.0998 | 69.05 | 71.4 | |
| MonoSDFOptimization=per-scene optimization, Priors=monocular geometric2026.05 | 0.125 | 63.17 | 70.39 | |
| 2DGSOptimization=per-scene optimization, Priors=prior-free2026.05 | 0.2181 | 48.7 | 49.03 |