Research map / 3D vision / Depth and stereo
Boost 3D Reconstruction using Diffusion-based Monocular Camera Calibration
Junyuan Deng, Wei Yin, Xiaoyang Guo, Qian Zhang, Xiaotao Hu, Weiqiang Ren, Xiao-Xiao Long, Ping Tan
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 Camera Image representation (with multi-res noise) (ablation) | diffusion camera calibration · GSV calibration ablation | mean FoV error (deg): 9.10 -> 2.54 | Improved |
| add multi-resolution noise (with Camera Image) (ablation) | diffusion camera calibration · GSV calibration ablation | mean FoV error (deg): 6.72 -> 2.54 | Improved |
| add camera image input to metric depth (ablation) | diffusion metric depth · NYU-v2 | delta1: 83.8 -> 85.8 | Improved |
| add camera image input to metric depth (ablation) | diffusion metric depth · KITTI | delta1: 87.8 -> 89.1 | Improved |
| add real data to training (ablation) | diffusion metric depth · NYU-v2 | delta1: 26.5 -> 85.8 | Improved |
| use one-step deterministic prediction instead of multi-step denoising (ablation) | diffusion metric depth · NYU-v2 | delta1: 77.1 -> 85.8 | Improved |
| train the VAE decoder instead of freezing it (ablation) | diffusion metric depth · NYU-v2 | delta1: 76.7 -> 85.8 | Improved |
| train the VAE decoder instead of freezing it (ablation) | diffusion metric depth · NYU-v2 | A.Rel: 48.0 -> 13.5 | Improved |
| add camera image to metric depth (ablation) | diffusion metric depth · iBims / DIODE indoor / DIODE outdoor | delta1: ibims 82.6 -> 88.7; Diode indoor 35.0 -> 50.1; Diode outdoor 25.2 -> 41.0 | Improved |
| add camera image to metric depth (ablation) | diffusion metric depth · NYU-v2 | SI_log: 8.0 -> 8.17 | No clear effect |
| add estimated intrinsics as cue to sparse-view reconstruction | DUSt3R · Relative distance error (4 in-the-wild scenes) | relative distance error: Sofa 1.67 -> 1.37; Car 0.87 -> 0.68; Pavilion 1.03 -> 0.68; StoneWall 1.43 -> 1.06 | Improved |
| add estimated intrinsics as cue to pose estimation | DUSt3R · pose estimation (20 images) | t_rel / r_rel: t_rel 1.17 -> 0.63 m; r_rel 5.02 -> 2.30 deg | 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.
- DiffCalib — DiffCalib: Reformulating Monocular Camera Calibration as Diffusion-Based Dense Incident Map Generation (AAAI 2025) · better on 2 of 2 shared results
- GeoCalib — GeoCalib: Learning Single-image Calibration with Geometric Optimization (ECCV 2024) · 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
- DiffCalib: Reformulating Monocular Camera Calibration as Diffusion-Based Dense Incident Map Generation (AAAI 2025 · similarity 0.92)
- Sparse3D: Distilling Multiview-Consistent Diffusion for Object Reconstruction from Sparse Views (AAAI 2024 · similarity 0.88)
- Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation (CVPR 2024 · similarity 0.88)
- Era3D: High-Resolution Multiview Diffusion using Efficient Row-wise Attention (NeurIPS 2024 · similarity 0.87)
- Cameras as Rays: Pose Estimation via Ray Diffusion (ICLR 2024 · similarity 0.87)
- UniK3D: Universal Camera Monocular 3D Estimation (CVPR 2025 · similarity 0.87)
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.
