Research map / Detection and segmentation / CNN features
Structured Modeling of Joint Deep Feature and Prediction Refinement for Salient Object Detection
Yingyue Xu, Dan Xu, Xiaopeng Hong, Wanli Ouyang, Rongrong Ji, Min Xu, Guoying Zhao
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 prediction-prediction message passing via jointly trained CRF (ablation) | Baseline · ECSSD | F-measure: 0.884 -> 0.921 | Improved |
| add feature-feature message passing (ablation) | Baseline · ECSSD | F-measure: 0.884 -> 0.904 | Improved |
| add feature-prediction messages to feature-feature and prediction-prediction (ablation) | CRF model · ECSSD | F-measure: 0.921 -> 0.928 | Improved |
| replace Dense-CRF post-processing with jointly trained CRF (ablation) | Baseline · ECSSD | F-measure: 0.902 -> 0.921 | Improved |
| joint training of CRF through backpropagation (backbone output) (ablation) | Baseline · ECSSD | F-measure: 0.884 -> 0.899 | Improved |
| emphasize similarity kernel sigma_alpha=10 sigma_beta=10 (ablation) | one CRF block at scale 5 · ECSSD | F-measure: 0.892 -> 0.910 | Improved |
| emphasize proximity kernel (sigma_beta=10, sigma_gamma=10) (ablation) | one CRF block at scale 5 · ECSSD | MAE: 0.071 -> 0.069 | Improved |
| change nu1 and nu2 weights (3,5 / 5,3) (ablation) | one CRF block at scale 5 · ECSSD | F-measure: 0.892 -> 0.892 | No clear effect |
| add CRF at scale 5 to the backbone (ablation) | CRF cascade · ECSSD | F-measure: 0.884 -> 0.921 | No clear effect |
| add vertical flipping augmentation (ablation) | CRF block · ECSSD | F-measure: 0.910 -> 0.905 | Hurt |
| add image noise augmentation (ablation) | CRF block · ECSSD | F-measure: 0.910 -> 0.896 | Hurt |
| add feature-prediction message passing (ablation) | CRF cascade · ECSSD | training loss | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Amulet — Amulet: Aggregating Multi-Level Convolutional Features for Salient Object Detection (ICCV 2017) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Recurrently Aggregating Deep Features for Salient Object Detection (AAAI 2018 · similarity 0.89)
- A Bi-Directional Message Passing Model for Salient Object Detection (CVPR 2018 · similarity 0.89)
- Salient Object Detection with Recurrent Fully Convolutional Networks (TPAMI 2018 · similarity 0.88)
- Cascaded Partial Decoder for Fast and Accurate Salient Object Detection (CVPR 2019 · similarity 0.88)
- A Stagewise Refinement Model for Detecting Salient Objects in Images (ICCV 2017 · similarity 0.88)
- Understanding and Visualizing Deep Visual Saliency Models (CVPR 2019 · 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.
