Research map / Detection and segmentation / CNN features
CapSal: Leveraging Captioning to Boost Semantics for Salient Object 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 |
|---|---|---|---|
| add captioning semantics to local-global saliency prediction (ablation) | CapSal LGPN · COCO-CapSal | Fβ: 0.821 -> 0.860 | Helped |
| add captioning semantics to local-global saliency prediction (ablation) | CapSal LGPN · DUTS-test | Fβ: 0.717 -> 0.756 | Helped |
| use local and global perception together instead of either module alone (ablation) | CapSal without captioning · COCO-CapSal | Fβ: 0.813 -> 0.821; 0.786 -> 0.821 | Helped |
| add joint training of ICN and LGPN (ablation) | CapSal · COCO-CapSal | Fβ: 0.843 -> 0.860 | Helped |
| add joint training of ICN and LGPN (ablation) | CapSal · DUTS-test | Fβ: 0.720 -> 0.756 | Helped |
| add textual attention to the caption embedding (ablation) | CapSal ICN · COCO-CapSal | Fβ: 0.837 -> 0.860 | Helped |
| add textual attention to the caption embedding (ablation) | CapSal ICN · DUTS-test | Fβ: 0.725 -> 0.756 | Helped |
| use ground-truth captions instead of generated captions (ablation) | CapSal LGPN · COCO-CapSal | Fβ: 0.860 -> 0.866 | Helped |
| train CapSal on DUTS-train rather than COCO-CapSal | CapSal · DUTS-test | Fβ: 0.756 -> 0.789 | Helped |
| train CapSal on COCO-CapSal rather than DUTS-train | CapSal · COCO-CapSal | Fβ: 0.815 -> 0.860 | Helped |
| use CapSal trained on DUTS-train instead of the strongest baseline | CapSal · PASCAL-S | MAE: 0.072 -> 0.064 | Helped |
| use CapSal trained on DUTS-train instead of the strongest baseline | CapSal · HKU-IS | Fβ: 0.895 -> 0.878 | Hurt |
Closest papers by meaning
- Beyond a Pre-Trained Object Detector: Cross-Modal Textual and Visual Context for Image Captioning (CVPR 2022 · similarity 0.89)
- Progressive Attention Guided Recurrent Network for Salient Object Detection (CVPR 2018 · similarity 0.89)
- Revisiting Salient Object Detection: Simultaneous Detection, Ranking, and Subitizing of Multiple Salient Objects (CVPR 2018 · similarity 0.89)
- Language-Guided Salient Object Ranking (CVPR 2025 · similarity 0.87)
- DetCLIPv3: Towards Versatile Generative Open-vocabulary Object Detection (CVPR 2024 · similarity 0.87)
- Benchmarking Large Vision-Language Models via Directed Scene Graph for Comprehensive Image Captioning (CVPR 2025 · 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.