Research map / Detection and segmentation / CNN features
Instance-Level Relative Saliency Ranking with Graph Reasoning
Nian Liu, Long Li, Wangbo Zhao, Junwei Han, Ling Shao
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 interaction relation graph G^r (ablation) | Baseline I · Proposed dataset test set | SA-SOR: 0.5147 -> 0.5339 | Improved |
| add local contrast graph G^l (ablation) | G^r model · Proposed dataset test set | SA-SOR: 0.5339 -> 0.5487 | Improved |
| add global contrast graph G^g (ablation) | G^r+G^l model · Proposed dataset test set | SA-SOR: 0.5487 -> 0.5568 | Improved |
| add person prior graph G^p (ablation) | G^r+G^l+G^g model · Proposed dataset test set | SA-SOR: 0.5568 -> 0.5616 | Improved |
| use graph reasoning instead of plain feature concatenation with same four features (ablation) | Baseline II · Proposed dataset test set | SA-SOR: 0.5419 -> 0.5616 | Improved |
| use K=8 subgraphs instead of K=16 (ablation) | G^r · Proposed dataset test set | SA-SOR: 0.5164 -> 0.5339 | Improved |
| use K=8 subgraphs instead of K=4 (ablation) | G^r · Proposed dataset test set | SA-SOR: 0.5160 -> 0.5339 | Improved |
| increase global context nodes from M=3 to M=7 (ablation) | G^g · Proposed dataset test set | SA-SOR: 0.5455 -> 0.5568 | Improved |
| increase global context nodes from M=7 to M=14 (ablation) | G^g · Proposed dataset test set | SA-SOR: 0.5568 -> 0.5568 | No clear effect |
| use ranking loss instead of Softmax with cross entropy (ablation) | Full model · Proposed dataset test set | SA-SOR: 0.5271 -> 0.5616 | Improved |
| add dynamic loss weight gamma=1 (ablation) | Ranking loss · Proposed dataset test set | SA-SOR: 0.5616 -> 0.5647 | Improved |
| increase gamma to 1.5 (ablation) | Ranking loss · Proposed dataset test set | SA-SOR: 0.5647 -> 0.5345 | Hurt |
Closest papers by meaning
- Revisiting Salient Object Detection: Simultaneous Detection, Ranking, and Subitizing of Multiple Salient Objects (CVPR 2018 · similarity 0.93)
- Salient Object Ranking With Position-Preserved Attention (ICCV 2021 · similarity 0.92)
- Language-Guided Salient Object Ranking (CVPR 2025 · similarity 0.89)
- A Motion-aware Spatio-temporal Graph for Video Salient Object Ranking (NeurIPS 2024 · similarity 0.89)
- Advancing Saliency Ranking with Human Fixations: Dataset Models and Benchmarks (CVPR 2024 · similarity 0.89)
- Unsupervised Salient Instance Detection (CVPR 2024 · similarity 0.88)
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.
