Research map / Video / Video segmentation and tracking
Unsupervised Video Object Segmentation with Motion-based Bilateral Networks
Siyang Li, Bryan Seybold, Alexey Vorobyov, Xuejing Lei, C.-C. Jay Kuo
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 objectness-based adaptive thresholding to the motion-based BNN (ablation) | Motion-based BNN · DAVIS 2016 val (no CRF) | J Mean: 53.8 -> 74.7 | Improved |
| add objectness thresholding to the BNN (F measure) (ablation) | Motion-based BNN · DAVIS 2016 val (no CRF) | F Mean: 50.1 -> 70.9 | Improved |
| add the embedding graph cut on top of Obj-BNN (ablation) | Obj-BNN · DAVIS 2016 val (no CRF) | J Mean: 74.7 -> 77.3 | Improved |
| use embedding similarity only as pairwise cost, versus none (ablation) | Embedding graph cut · DAVIS 2016 val (no CRF) | J Mean: 74.7 -> 74.3 | No clear effect |
| use motion similarity only as pairwise cost (ablation) | Embedding graph cut · DAVIS 2016 val (no CRF) | J Mean: 74.7 -> 74.8 | No clear effect |
| combine embedding and motion similarity in the pairwise cost (ablation) | Embedding graph cut without temporal edges · DAVIS 2016 val (no CRF) | J Mean: 74.3 -> 75.7 | Improved |
| add sparse inter-frame temporal edges (ablation) | Embedding graph cut · DAVIS 2016 val (no CRF) | J Mean: 75.7 -> 76.2 | Improved |
| use dense instead of sparse temporal edges (ablation) | Embedding graph cut · DAVIS 2016 val (no CRF) | J Mean: 76.2 -> 77.3 | Improved |
| add CRF refinement to the full method (ablation) | Full method · DAVIS 2016 val | J Mean: 77.3 -> 80.4 | Improved |
| widen causal graph window from W=0 to W=1 (current frame labels) (ablation) | Embedding graph cut · DAVIS 2016 val (no CRF) | J Mean: 75.3 -> 76.8 | Improved |
| increase causal window beyond W=1 to 5 or 10 (ablation) | Embedding graph cut · DAVIS 2016 val (no CRF) | J Mean: 76.8 -> 76.9 | No clear effect |
| propagate labels from previous and following frames seeds in addition to current (ablation) | Label propagation (acausal W=1) · DAVIS 2016 val (no CRF) | J Mean: 77.0 -> 77.3 | Improved |
Closest papers by meaning
- Unsupervised Multi-Object Segmentation by Predicting Probable Motion Patterns (NeurIPS 2022 · similarity 0.90)
- Unsupervised Object Segmentation in Video by Efficient Selection of Highly Probable Positive Features (ICCV 2017 · similarity 0.90)
- Unsupervised Video Object Segmentation using Motion Saliency-Guided Spatio-Temporal Propagation (ECCV 2018 · similarity 0.89)
- Instance Embedding Transfer to Unsupervised Video Object Segmentation (CVPR 2018 · similarity 0.89)
- Deep Transport Network for Unsupervised Video Object Segmentation (ICCV 2021 · similarity 0.89)
- unMORE: Unsupervised Multi-Object Segmentation via Center-Boundary Reasoning (ICML 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.
