Research map / Video / Video segmentation and tracking
Structure Matters: Revisiting Boundary Refinement in Video Object Segmentation
Guanyi Qin, Ziyue Wang, Daiyun Shen, Haofeng Liu, Hantao Zhou, Junde Wu, Runze Hu, Yueming Jin
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 Canny-edge enhanced features, structure decoder, object memory fusion and evidential loss (OASIS) (ablation) | Cutie-style baseline · DAVIS-17 val (DAVIS-only training) | J&F: 84.8 -> 85.6 | Improved |
| add Canny edge feature integration (+CaE) (ablation) | baseline · DAVIS-17 val | J&F: 84.8 -> 84.9 | No clear effect |
| add structure decoder on top of CaE (ablation) | +CaE · DAVIS-17 test-dev | J&F: 80.0 -> 80.3 | Improved |
| add object memory feature fusion in the structure decoder (ablation) | +CaE+SD · DAVIS-17 test-dev | J&F: 80.3 -> 80.8 | Improved |
| add evidential learning loss on top of structure refinement (ablation) | +CaE+SD* · DAVIS-17 test-dev | J&F: 80.8 -> 81.0 | Improved |
| raise beta and epsilon beyond optimum (ablation) | OASIS · DAVIS-17 val | J&F: beta 85.7 (2.0) -> 85.0 (5.0); epsilon 85.6 (0.5) -> 85.0 (2.0) | Hurt |
| add 2M-parameter projection network instead of structure module (ablation) | baseline · DAVIS-17 val | J&F: 84.8 -> 85.1 (OASIS 85.6) | No clear effect |
| add structure refinement and EDL (OASIS) versus Cutie | Cutie · DAVIS-17 val | J&F: 86.8 -> 88.3 | Improved |
| add structure refinement module (ablation) | baseline · DAVIS-17 val | FPS: 52 -> 48 | Hurt |
Closest papers by meaning
- Video Object Segmentation With Dynamic Memory Networks and Adaptive Object Alignment (ICCV 2021 · similarity 0.88)
- Video Object Segmentation without Temporal Information (TPAMI 2018 · similarity 0.88)
- From ViT Features to Training-free Video Object Segmentation via Streaming-data Mixture Models (NeurIPS 2023 · similarity 0.88)
- Video Object Segmentation with Adaptive Feature Bank and Uncertain-Region Refinement (NeurIPS 2020 · similarity 0.87)
- CapsuleVOS: Semi-Supervised Video Object Segmentation Using Capsule Routing (ICCV 2019 · similarity 0.87)
- Using Diffusion Priors for Video Amodal Segmentation (CVPR 2025 · 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.
