Research map / Video / Video segmentation and tracking
A Multilayer-Based Framework for Online Background Subtraction With Freely Moving Cameras
Yizhe Zhu, Ahmed Elgammal
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 |
|---|---|---|---|
| model each foreground object in its own layer (multilayer vs best prior method) | MLBS · Hopkins + CBD overall | F-score (%): 77.73 -> 83.66 | Improved |
| use separate layers when scene has one moving object | MLBS · Hopkins, 1 object | F-score (%): 80.70 -> 84.05 | Improved |
| use separate layers when scene has two moving objects | MLBS · Hopkins, 2 objects | F-score (%): 83.19 -> 91.14 | Improved |
| use separate layers when scene has three or more moving objects | MLBS · Hopkins, 3+ objects | F-score (%): 66.26 -> 74.99 | Improved |
| replace online MLBS with offline SLT for multi-foreground segmentation | MLBS · Hopkins multi-foreground, all frames | F-score (%): 82.00 -> 86.16 | Improved |
| replace naive baseline with MLBS for multi-foreground segmentation | MLBS · Hopkins multi-foreground, all frames | F-score (%): 64.35 -> 86.16 | Improved |
| evaluate on all frames rather than only the first 10 frames | MLBS · Hopkins multi-foreground | F-score (%): 85.06 -> 86.16 | Improved |
| use MLBS instead of GBSSP on a sequence with objects occluding then separating | MLBS · Hopkins cars3 | F-score (%): 72.94 -> 93.16 | Improved |
| use MLBS instead of GBSSP on forest sequence | MLBS · CBD forest | F-score (%): 91.44 -> 72.20 | Hurt |
| use MLBS instead of FOF on traffic sequence | MLBS · CBD traffic | F-score (%): 71.24 -> 48.19 | Hurt |
| approximate motion uncertainty by a Dirac delta when propagating appearance (ablation) | MLBS · model propagation | performance | No clear effect |
Closest papers by meaning
- Robust Online Matrix Factorization for Dynamic Background Subtraction (TPAMI 2017 · similarity 0.88)
- Online Mutual Foreground Segmentation for Multispectral Stereo Videos (IJCV 2019 · similarity 0.86)
- Foreground Segmentation with Tree-Structured Sparse RPCA (TPAMI 2017 · similarity 0.85)
- Spatiotemporal GMM for Background Subtraction with Superpixel Hierarchy (TPAMI 2017 · similarity 0.84)
- Video Segmentation via Multiple Granularity Analysis (CVPR 2017 · similarity 0.84)
- 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.
