Research map / Detection and segmentation / Tracking
FMODetect: Robust Detection of Fast Moving Objects
Denys Rozumnyi, Jiří Matas, Filip Šroubek, Marc Pollefeys, Martin R. Oswald
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 energy-minimization deblurring to refine network trajectories (ablation) | FMODetect · TbD dataset | TIoU: 0.519 -> 0.715 | Improved |
| add energy-minimization deblurring to refine network trajectories (ablation) | FMODetect · TbD dataset | recall: 0.97 -> 0.97 | No clear effect |
| add energy-minimization deblurring to refine network trajectories (ablation) | FMODetect · TbD dataset | runtime: 20 fps -> 0.4 fps | Hurt |
| add non-causal post-processing (ablation) | FMODetect + deblurring · TbD dataset | TIoU: 0.715 -> 0.781 | Improved |
| add non-causal post-processing (ablation) | FMODetect + deblurring · TbD dataset | recall: 0.97 -> 0.99 | Improved |
| add binary FMO predictor output to the matting and fitting network (ablation) | FMODetect · TbD dataset | false positives: removes approximately 70% of false positives | Improved |
| split deblatting into matting by network then deblurring (removing background first) | FMODetect deblurring · TbD dataset | deblurring speed: twice as fast; 10 times faster without H | Improved |
| use FMODetect instead of TbD | FMODetect · FMO dataset | recall (%): 41.1 -> 48.8 | Improved |
| use FMODetect instead of TbD | FMODetect · TbD-3D dataset | precision (%): 98.2 -> 87.9 | Hurt |
| use FMODetect instead of TbD | FMODetect · TbD-3D dataset | recall (%): 69.8 -> 83.5 | Improved |
| use FMODetect instead of TbD | FMODetect · YouTube trick shots | recall (%): 27.3 -> 90.1 | Improved |
| use FMODetect instead of TbD | FMODetect · Falling objects dataset | precision (%): 92.2 -> 98.9 | Improved |
Closest papers by meaning
- Tracking by Deblatting (IJCV 2021 · similarity 0.90)
- Sub-Frame Appearance and 6D Pose Estimation of Fast Moving Objects (CVPR 2020 · similarity 0.86)
- DeFMO: Deblurring and Shape Recovery of Fast Moving Objects (CVPR 2021 · similarity 0.86)
- Separation for Better Integration: Disentangling Edge and Motion in Event-based Deblurring (ICCV 2025 · similarity 0.85)
- The World of Fast Moving Objects (CVPR 2017 · similarity 0.85)
- DiffusionTrack: Diffusion Model for Multi-Object Tracking (AAAI 2024 · similarity 0.84)
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.
