Research map / 3D vision / Depth and stereo
BLASt3R: Bundle Adjustment of Any Image Set with Multi-View Matching and Monocular priors
Vincent Leroy, Philippe Weinzaepfel, Lojze Zust, Yohann Cabon, Jerome Revaud
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 multiple corrective depth channels (nz=16 instead of nz=1) (ablation) | BLASt3R · Monocular depth, 10-dataset average | Rel: 8.65 -> 4.87 | Improved |
| add multiple corrective depth channels (nz=16 instead of nz=1) (ablation) | BLASt3R · Monocular depth, 10-dataset average | delta1: 90.3 -> 96.2 | Improved |
| add multiple depth channels (nz=1 -> full) (ablation) | BLASt3R · ETH3D dense reconstruction | Accuracy mean: 0.104 -> 0.069 | Improved |
| add multiple depth channels (nz=1 -> full) (ablation) | BLASt3R · 7-Scenes dense reconstruction | Accuracy mean: 0.029 -> 0.021 | Improved |
| add multiple depth channels (nz=1 -> full) (ablation) | BLASt3R · ETH3D dense reconstruction | Normal consistency mean: 0.835 -> 0.881 | Improved |
| add multiple depth channels (nz=1 -> full) (ablation) | BLASt3R · ETH3D dense reconstruction | Completeness mean: 0.045 -> 0.055 | Hurt |
| use depth-regularized BA instead of COLMAP BA (ablation) | BLASt3R · Co3Dv2 (10 frames) | mAA@30: 86.3 -> 88.2 | Improved |
| use depth-regularized BA instead of COLMAP BA (ablation) | BLASt3R · RealEstate10K (10 frames) | mAA@30: 80.2 -> 85.2 | Improved |
| use depth-regularized BA instead of COLMAP BA (ablation) | BLASt3R · ETH3D-MVS | mAA@30: 98.8 -> 98.5 | Hurt |
| use depth-regularized BA instead of COLMAP BA (ablation) | BLASt3R · Tanks&Temples (full sequences) | mAA@30: 78.5 -> 74.1 | Hurt |
| use calibrated intrinsics instead of uncalibrated | BLASt3R · TUM-RGBD fr1 | ATE RMSE (cm), avg: 3.0 -> 2.5 | Improved |
| disable focal optimization in initial BA iterations | BLASt3R · RealEstate10K (10 frames) | mAA@30: 85.2 -> 88.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.
- MUSt3R-512 — MUSt3R: Multi-view Network for Stereo 3D Reconstruction (CVPR 2025) · better on 1 of 1 shared result
- MoGe — MoGe: Unlocking Accurate Monocular Geometry Estimation for Open-Domain Images with Optimal Training Supervision (CVPR 2025) · better on 0 of 1 shared result
- DROID-SLAM — DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Cameras (NeurIPS 2021) · better on 1 of 1 shared result
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
- Dynamic Visual SLAM using a General 3D Prior (CVPR 2026 · similarity 0.89)
- BA-T: An Iterative Transformer for Two-View Bundle Adjustment (NeurIPS 2026 · similarity 0.88)
- Light3R-SfM: Towards Feed-forward Structure-from-Motion (CVPR 2025 · similarity 0.87)
- MASt3R-SLAM: Real-Time Dense SLAM with 3D Reconstruction Priors (CVPR 2025 · similarity 0.87)
- SLAM3R: Real-Time Dense Scene Reconstruction from Monocular RGB Videos (CVPR 2025 · similarity 0.86)
- ProDyG: Progressive Dynamic Scene Reconstruction via Gaussian Splatting from Monocular Videos (NeurIPS 2025 · similarity 0.86)
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.
