Research map / Detection and segmentation / CNN features
NUN: Nested Unfolding Network for Real-World Concealed Object Segmentation
Chunming He, Rihan Zhang, Longxiang Tang, Dingming Zhang, Bojian Zhang, Fengyang Xiao, Jingjia Feng, Sina Farsiu
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 DeRUN restoration unfolding network (ablation) | SODUN · COD10K combined degradation | F_beta: 0.362 -> 0.406 | Improved |
| add bi-directional unfolding interaction (BUI) (ablation) | SODUN + DeRUN · COD10K combined degradation | F_beta: 0.406 -> 0.448 | Improved |
| add SODUN background estimation (ablation) | SODUN- · COD10K combined degradation | F_beta: 0.328 -> 0.362 | Improved |
| use DA-CLIP explicit degradation prior instead of implicit prior (ablation) | DeRUN · COD10K combined degradation | F_beta: 0.422 -> 0.448 | Improved |
| add segmentation cues X2 in DeRUN (ablation) | DeRUN · COD10K combined degradation | F_beta: 0.430 -> 0.448 | Improved |
| add IQA-based selection of best restoration (ablation) | BUI · COD10K combined degradation | F_beta: 0.421 -> 0.448 | Improved |
| add cross-stage consistency loss (ablation) | BUI · COD10K combined degradation | F_beta: 0.428 -> 0.448 | Improved |
| use simple 3-layer network instead of complex transformer for background estimation (ablation) | SODUN · COD10K combined degradation | F_beta: 0.447 -> 0.448 | No clear effect |
| NUN vs RUN under combined degradation | NUN · COD10K combined degradation (PVT V2) | F_beta: 0.429 -> 0.732 | Improved |
| NUN vs RUN on clean COD10K | NUN · COD10K clean (ResNet50) | F_beta: 0.747 -> 0.759 | Improved |
| NUN adds restoration/segmentation nesting at compute cost | NUN · COD efficiency (ResNet50) | FPS: 22.75 -> 20.10 | Hurt |
| integrate NUN framework into existing models instead of pre-restoration or bi-level optimization | FEDER · COD10K combined degradation | F_beta: 0.322 -> 0.416 | Improved |
Closest papers by meaning
- RUN: Reversible Unfolding Network for Concealed Object Segmentation (ICML 2025 · similarity 0.88)
- Segment Anything without Supervision (NeurIPS 2024 · similarity 0.85)
- Hierarchy-Agnostic Unsupervised Segmentation: Parsing Semantic Image Structure (NeurIPS 2024 · similarity 0.85)
- UniFunc3D: Unified Active Spatial-Temporal Grounding for 3D Affordance Segmentation (NeurIPS 2026 · similarity 0.85)
- Generative Omnimatte: Learning to Decompose Video into Layers (CVPR 2025 · similarity 0.85)
- SSUN-Net: Spatial-Spectral Prior-Aware Unfolding Network for Pan-Sharpening (AAAI 2025 · similarity 0.85)
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.
