Research map / 3D vision / Face and shape
Shape from Blur: Recovering Textured 3D Shape and Motion of Fast Moving Objects
Denys Rozumnyi, Martin R. Oswald, Vittorio Ferrari, Marc Pollefeys
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 silhouette consistency loss L_S (ablation) | SfB · Falling Objects | TIoU: 0.341 -> 0.693 | Improved |
| Add silhouette consistency loss L_S (ablation) | SfB · Falling Objects | PSNR: 17.94 -> 26.30 | Improved |
| Use DeFMO-mask L_S instead of background-subtraction L_D (ablation) | SfB · Falling Objects | PSNR: 20.71 -> 26.30 | Improved |
| Add texture loss L_F to L_S (ablation) | SfB · Falling Objects | PSNR: 26.30 -> 25.64 | Hurt |
| Add Laplacian shape regularization L_L (ablation) | SfB · Falling Objects | TIoU: 0.676 -> 0.693 | Improved |
| Add Laplacian shape regularization L_L (ablation) | SfB · Synthetic 3D dataset | translation error: 0.462 -> 0.305 | Improved |
| Add texture smoothness L_T (ablation) | SfB · Falling Objects | PSNR: 24.98 -> 26.30 | Improved |
| Use sphere prototype instead of torus (ablation) | SfB · Falling Objects | translation error (synthetic): 1.741 -> 0.305 | Improved |
| Model rotation (sphere with rotation vs no rotation) (ablation) | SfB · Synthetic 3D dataset | translation error: 0.835 -> 0.305 | Improved |
| Increase prototypes from 1 to 3 (ablation) | SfB · Falling Objects | TIoU: 0.693 -> 0.701 | Improved |
| Add 4 distractor prototypes to 3 prototypes (ablation) | SfB · Falling Objects | TIoU: 0.701 -> 0.696 | No clear effect |
| Increase time steps from 1x8 to 5x8 (ablation) | SfB · Falling Objects | PSNR: 26.30 -> 26.39 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DeFMO — DeFMO: Deblurring and Shape Recovery of Fast Moving Objects (CVPR 2021) · better on 9 of 9 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
- Motion-From-Blur: 3D Shape and Motion Estimation of Motion-Blurred Objects in Videos (CVPR 2022 · similarity 0.91)
- DeFMO: Deblurring and Shape Recovery of Fast Moving Objects (CVPR 2021 · similarity 0.88)
- SF3D: Stable Fast 3D Mesh Reconstruction with UV-unwrapping and Illumination Disentanglement (CVPR 2025 · similarity 0.87)
- GS-Blur: A 3D Scene-Based Dataset for Realistic Image Deblurring (NeurIPS 2024 · similarity 0.86)
- BARD-GS: Blur-Aware Reconstruction of Dynamic Scenes via Gaussian Splatting (CVPR 2025 · similarity 0.86)
- Coherent 3D Scene Diffusion From a Single RGB Image (NeurIPS 2024 · 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.
