Research map / 3D vision / Depth and stereo
CAM3R: Camera-Agnostic Model for 3D Reconstruction
Namitha Guruprasad, Abhay Kumar Yadav, Cheng Peng, Rama Chellappa
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add heterogeneous camera pairs in second training phase (ablation) | CAM3R · 2D3DS two-view | RTA@15: 56.8 -> 94.3 | Improved |
| add heterogeneous camera pairs in second training phase (ablation) | CAM3R · 360Loc two-view | RTA@15: 54.7 -> 91.0 | Improved |
| add heterogeneous camera pairs in second training phase (ablation) | CAM3R · MegaDepth two-view | RRA@15: 97.2 -> 96.8 | Hurt |
| replace DUSt3R global alignment with ray-aware global alignment (ablation) | CAM3R · 2D3DS multi-view | mAA@30: 38.8 -> 73.5 | Improved |
| replace DUSt3R global alignment with ray-aware global alignment (ablation) | CAM3R · 360Loc multi-view | ATE RMSE: 4.5 -> 2.7 | Improved |
| replace DUSt3R global alignment with ray-aware global alignment (ablation) | CAM3R · MegaDepth multi-view | mAA@30: 68.5 -> 87.4 | Improved |
| decouple rays and radial distance with Ray Module versus naive DUSt3R fine-tuning (ablation) | CAM3R · 360Loc two-view | RTA@15: 9.2 -> 91.0 | Improved |
| naively fine-tune DUSt3R on distorted data (ablation) | DUSt3R · MegaDepth two-view | RTA@15: 80.8 -> 72.5 | Hurt |
| CAM3R vs pi3 on pinhole MegaDepth multi-view and CO3Dv2 RTA | CAM3R · CO3Dv2 / MegaDepth multi-view | ATE RMSE: 0.7 vs 1.1 on CO3Dv2; 0.6 vs 0.8 on MegaDepth | Hurt |
| CAM3R vs best pinhole baseline in zero-shot CO3Dv2 two-view | CAM3R · CO3Dv2 two-view | RTA@15: 43.1 -> 88.2 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- VGGT — VGGT: Visual Geometry Grounded Transformer (CVPR 2025) · better on 7 of 7 shared results
- DUSt3R — DUSt3R: Geometric 3D Vision Made Easy (CVPR 2024) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DUSt3R: Geometric 3D Vision Made Easy (CVPR 2024 · similarity 0.89)
- PF3plat: Pose-Free Feed-Forward 3D Gaussian Splatting for Novel View Synthesis (ICML 2025 · similarity 0.88)
- UniK3D: Universal Camera Monocular 3D Estimation (CVPR 2025 · similarity 0.88)
- Fin3R: Fine-tuning Feed-forward 3D Reconstruction Models via Monocular Knowledge Distillation (NeurIPS 2025 · similarity 0.88)
- SPARS3R: Semantic Prior Alignment and Regularization for Sparse 3D Reconstruction (CVPR 2025 · similarity 0.88)
- MUSt3R: Multi-view Network for Stereo 3D Reconstruction (CVPR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
