Research map / Detection and segmentation / CNN features
PiCANet: Learning Pixel-Wise Contextual Attention for Saliency Detection
Read the paperCheck 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 hierarchical global and local pixel-wise contextual attention to U-Net (ablation) | U-Net saliency model · DUT-O | weighted F-measure: 0.651 -> 0.691 | Helped |
| add hierarchical global and local pixel-wise contextual attention to U-Net (ablation) | U-Net saliency model · DUTS-TE | weighted F-measure: 0.715 -> 0.748 | Helped |
| add global PiCANets in D7 and D5 (ablation) | U-Net saliency model · DUT-O | weighted F-measure: 0.651 -> 0.662 | Helped |
| add global PiCANets in D7 and D5 (ablation) | U-Net saliency model · DUTS-TE | weighted F-measure: 0.715 -> 0.724 | Helped |
| replace PiCANets with max-pooling (ablation) | U-Net with global and local context aggregation · DUT-O | weighted F-measure: 0.691 -> 0.671 | Hurt |
| replace PiCANets with average-pooling (ablation) | U-Net with global and local context aggregation · DUT-O | weighted F-measure: 0.691 -> 0.670 | Hurt |
| use local PiCANets only, omitting global attention (ablation) | U-Net saliency model · DUT-O | weighted F-measure: 0.691 -> 0.680 | Hurt |
| use local PiCANets only, omitting global attention (ablation) | U-Net saliency model · DUTS-TE | weighted F-measure: 0.748 -> 0.738 | Hurt |
| move global attention from D7 and D5 to D4 (ablation) | U-Net saliency model · DUT-O | weighted F-measure: 0.691 -> 0.688 | No clear effect |
| move global attention from D7 and D5 to D4 (ablation) | U-Net saliency model · DUTS-TE | weighted F-measure: 0.748 -> 0.747 | No clear effect |
| add a global PiCANet in D7 and use local PiCANets in D5–D2 (ablation) | U-Net saliency model · DUT-O | weighted F-measure: 0.691 -> 0.690 | No clear effect |
| add a global PiCANet in D7 and use local PiCANets in D5–D2 (ablation) | U-Net saliency model · DUTS-TE | weighted F-measure: 0.748 -> 0.744 | No clear effect |
Closest papers by meaning
- Progressive Attention Guided Recurrent Network for Salient Object Detection (CVPR 2018 · similarity 0.89)
- Detect Globally, Refine Locally: A Novel Approach to Saliency Detection (CVPR 2018 · similarity 0.89)
- Recurrent Attentional Networks for Saliency Detection (CVPR 2016 · similarity 0.88)
- Multi-Type Self-Attention Guided Degraded Saliency Detection (AAAI 2020 · similarity 0.88)
- Samba: A Unified Mamba-based Framework for General Salient Object Detection (CVPR 2025 · similarity 0.87)
- Explainable Saliency: Articulating Reasoning with Contextual Prioritization (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.