Research map / Detection and segmentation / CNN features
A Stagewise Refinement Model for Detecting Salient Objects in Images
Tiantian Wang, Ali Borji, Lihe Zhang, Pingping Zhang, Huchuan Lu
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 a second refinement stage (ablation) | 1-stage baseline · ECSSD | F-measure: 0.843 -> 0.867 | Improved |
| add a second refinement stage (ablation) | 1-stage baseline · DUTS | F-measure: 0.682 -> 0.706 | Improved |
| add pyramid pooling module to 2-stage (ablation) | 2-stage · DUTS | F-measure: 0.706 -> 0.734 | Improved |
| add pyramid pooling module to 2-stage (ablation) | 2-stage · ECSSD | F-measure: 0.867 -> 0.878 | Improved |
| add pyramid pooling module to 3-stage (ablation) | 3-stage · ECSSD | F-measure: 0.879 -> 0.892 | Improved |
| add pyramid pooling module to 3-stage (ablation) | 3-stage · DUTS | F-measure: 0.735 -> 0.757 | Improved |
| go from 3 to 4 stages with ppm (ablation) | 3-stage+ppm · ECSSD | F-measure: 0.892 -> 0.882 | Hurt |
| go from 3 to 4 stages with ppm (ablation) | 3-stage+ppm · DUTS | F-measure: 0.757 -> 0.742 | No clear effect |
| use stage-wise refinement instead of ResNet-50 FCN with skip connections (ablation) | ResNet-50 FCN · ECSSD | F-measure: 0.864 -> 0.892 | Improved |
| use stage-wise refinement instead of ResNet-50 FCN with skip connections (ablation) | ResNet-50 FCN · DUTS | F-measure: 0.689 -> 0.757 | Improved |
| add auxiliary losses on intermediate stages (ablation) | 3-stage network · DUT-OMRON and DUTS | F-measure | Improved |
| train each stage independently first then jointly instead of direct end-to-end training (ablation) | 3-stage network · DUT-OMRON and DUTS | F-measure | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- RFCN — R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016) · better on 6 of 6 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
- Detect Globally, Refine Locally: A Novel Approach to Saliency Detection (CVPR 2018 · similarity 0.92)
- Salient Object Detection with Recurrent Fully Convolutional Networks (TPAMI 2018 · similarity 0.91)
- Salient Object Detection With Pyramid Attention and Salient Edges (CVPR 2019 · similarity 0.91)
- Recurrently Aggregating Deep Features for Salient Object Detection (AAAI 2018 · similarity 0.91)
- BASNet: Boundary-Aware Salient Object Detection (CVPR 2019 · similarity 0.90)
- Instance-Level Salient Object Segmentation (CVPR 2017 · similarity 0.90)
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.
