Research map / Detection and segmentation / CNN features
Saliency Detection via Combining Region-Level and Pixel-Level Predictions with CNNs
Youbao Tang, Xiangqian Wu
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 joint training of pixel-level CNN and fusion CNN (ablation) | sep-CRPSD · SED | weighted F-measure: 0.8109 -> 0.8292 | Improved |
| add joint training of pixel-level CNN and fusion CNN (ablation) | sep-CRPSD · ECSSD | weighted F-measure: 0.8249 -> 0.8485 | Improved |
| add joint training of pixel-level CNN and fusion CNN (ablation) | sep-CRPSD · PASCAL-S | weighted F-measure: 0.7621 -> 0.7761 | Improved |
| add joint training of pixel-level CNN and fusion CNN (ablation) | sep-CRPSD · HKU-IS | weighted F-measure: 0.7942 -> 0.8209 | Improved |
| fuse region-level saliency and original image with pixel-level map via fusion CNN (ablation) | PSD · SED | weighted F-measure: 0.7590 -> 0.8292 | Improved |
| fuse region-level saliency and original image with pixel-level map via fusion CNN (ablation) | PSD · ECSSD | weighted F-measure: 0.7572 -> 0.8485 | Improved |
| fuse region-level saliency and original image with pixel-level map via fusion CNN (ablation) | PSD · PASCAL-S | weighted F-measure: 0.7113 -> 0.7761 | Improved |
| fuse region-level saliency and original image with pixel-level map via fusion CNN (ablation) | PSD · HKU-IS | weighted F-measure: 0.7371 -> 0.8209 | Improved |
| fuse pixel-level saliency and original image with region-level map via fusion CNN (ablation) | RSD · PASCAL-S | weighted F-measure: 0.6195 -> 0.7761 | Improved |
| train for saliency instead of using semantic segmentation model directly | pre-DML · ECSSD | weighted F-measure: 0.6530 -> 0.7482 | Improved |
| train for saliency instead of using semantic segmentation model directly | pre-DML · PASCAL-S | weighted F-measure: 0.7322 -> 0.6948 | Hurt |
| adaptive regions and region-centered context windows instead of superpixels (RSD vs MC) | MC · PASCAL-S | weighted F-measure: 0.6064 -> 0.6195 | Improved |
Closest papers by meaning
- Deep Saliency With Encoded Low Level Distance Map and High Level Features (CVPR 2016 · similarity 0.90)
- Deep Level Sets for Salient Object Detection (CVPR 2017 · similarity 0.90)
- ELD-Net: An Efficient Deep Learning Architecture for Accurate Saliency Detection (TPAMI 2017 · similarity 0.89)
- Non-Local Deep Features for Salient Object Detection (CVPR 2017 · similarity 0.88)
- Instance-Level Salient Object Segmentation (CVPR 2017 · similarity 0.88)
- Saliency Hierarchy Modeling via Generative Kernels for Salient Object Detection (ECCV 2022 · 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.
