Research map / Detection and segmentation / Weakly supervised detection and segmentation
Multi-Source Weak Supervision for Saliency Detection
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Combine category-label and caption supervision by averaging separately trained network outputs (ablation) | CNet and PNet · ECSSD | max Fβ: 0.730 -> 0.762 | Helped |
| Add attention transfer loss to jointly train CNet and PNet (ablation) | CNet + PNet · ECSSD | max Fβ: 0.762 -> 0.786 | Helped |
| Add unlabelled data with attention coherence loss (ablation) | CNet + PNet with attention transfer · ECSSD | max Fβ: 0.786 -> 0.820 | Helped |
| Train SNet on pseudo labels generated by CNet and PNet (ablation) | CNet + PNet with attention transfer and coherence · ECSSD | max Fβ: 0.820 -> 0.878 | Helped |
| Use the full multi-source weak-supervision framework instead of WSS | Saliency detection · ECSSD, PASCAL-S, SOD, MSRA5K, DUT-OMRON | max Fβ: ECSSD 0.856 -> 0.878; PASCAL-S 0.778 -> 0.790; SOD 0.780 -> 0.799; MSRA5K 0.877 -> 0.890; DUT-OMRON 0.687 -> 0.718 | Helped |
| Use the full multi-source weak-supervision framework instead of fully supervised DS | Saliency detection · ECSSD and DUT-OMRON | max Fβ: ECSSD 0.882 vs 0.878; DUT-OMRON 0.739 vs 0.718 | Hurt |
| Use the full multi-source weak-supervision framework instead of fully supervised baselines | Saliency detection · PASCAL-S and SOD | max Fβ: PASCAL-S 0.768 -> 0.790; SOD 0.786 -> 0.799 | Helped |
Closest papers by meaning
- Structure-Consistent Weakly Supervised Salient Object Detection with Local Saliency Coherence (AAAI 2021 · similarity 0.90)
- Learning to Detect Salient Objects With Image-Level Supervision (CVPR 2017 · similarity 0.90)
- Joint Semantic Mining for Weakly Supervised RGB-D Salient Object Detection (NeurIPS 2021 · similarity 0.90)
- Railroad Is Not a Train: Saliency As Pseudo-Pixel Supervision for Weakly Supervised Semantic Segmentation (CVPR 2021 · similarity 0.90)
- Weakly-Supervised Salient Object Detection Using Point Supervision (AAAI 2022 · similarity 0.89)
- Unsupervised Salient Instance Detection (CVPR 2024 · 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.