Research map / Video / Video segmentation and tracking
Video Object Segmentation by Learning Location-Sensitive Embeddings
Hai Ci, Chunyu Wang, Yizhou Wang
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 zooming-in on estimated object box (ablation) | FP · DAVIS val (first-frame box) | J mean: 70.5 -> 76.2 | Improved |
| jointly train location-sensitive embedding branch with foreground branch (ablation) | FP · DAVIS val (first-frame box) | J mean: 76.2 -> 80.0 | Improved |
| jointly train location-sensitive embedding branch (ablation) | FP · DAVIS val (first-frame box) | F mean: 75.0 -> 79.1 | Improved |
| fuse embedding-based foreground map with foreground prediction (ablation) | FPLSE · DAVIS val (first-frame box) | J mean: 80.0 -> 81.0 | Improved |
| add per-video model adaptation with pseudo-labels (ablation) | Ours (static) · DAVIS val (first-frame box) | J mean: 80.0 -> 81.0 | Improved |
| add per-video model adaptation with pseudo-labels (ablation) | Ours (static) · DAVIS val (first-frame box) | F recall: 87.0 -> 89.1 | Improved |
| add CRF post-processing | Ours · DAVIS val (first-frame box) | J mean: 81.0 -> 82.9 | Improved |
| add CRF post-processing | Ours · DAVIS val (first-frame box) | F mean: 80.3 -> 80.1 | No clear effect |
| add location-sensitive embeddings to foreground prediction (GT boxes) (ablation) | FP · DAVIS val (GT boxes all frames) | J mean: 80.4 -> 80.9 | Improved |
| replace GrabCut with learned foreground prediction network | GrabCut · DAVIS val (GT boxes all frames) | J mean: 61.5 -> 80.4 | Improved |
| add location-sensitive embeddings to foreground prediction (GT boxes) (ablation) | FP · DAVIS val (GT boxes all frames) | F mean: 80.2 -> 80.8 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- MaskRNN — MaskRNN: Instance Level Video Object Segmentation (NeurIPS 2017) · better on 0 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
- Video Object Segmentation Using Space-Time Memory Networks (ICCV 2019 · similarity 0.90)
- Learning Spatial-Semantic Features for Robust Video Object Segmentation (ICLR 2025 · similarity 0.89)
- Unsupervised Object Segmentation in Video by Efficient Selection of Highly Probable Positive Features (ICCV 2017 · similarity 0.89)
- Video Object Segmentation With Dynamic Memory Networks and Adaptive Object Alignment (ICCV 2021 · similarity 0.89)
- Video Object Segmentation Using Global and Instance Embedding Learning (CVPR 2021 · similarity 0.89)
- Learning Video Object Segmentation With Visual Memory (ICCV 2017 · similarity 0.89)
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.
