Research map / Detection and segmentation / CNN features
Contour Knowledge Transfer for Salient Object Detection
Xin Li, Fan Yang, Hong Cheng, Wei Liu, Dinggang Shen
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 cross-domain connections (CDC) between contour and saliency branches (ablation) | C2S-Net · ECSSD | weighted F-measure: 0.793 -> 0.812 | Improved |
| add cross-domain connections (CDC) between contour and saliency branches (ablation) | C2S-Net · ECSSD | MAE: 0.103 -> 0.081 | Improved |
| use contour-to-saliency transferring (CTS) instead of averaging proposals (AVG-P) (ablation) | C2S-Net + CDC · ECSSD | weighted F-measure: 0.665 -> 0.743 | Improved |
| use contour-to-saliency transferring (CTS) instead of winner-take-all (WTA) (ablation) | C2S-Net + CDC · ECSSD | weighted F-measure: 0.732 -> 0.743 | Improved |
| replace joint training (SCJ) with alternating training (AT(3)) (ablation) | C2S-Net + CDC + CTS · ECSSD | weighted F-measure: 0.759 -> 0.837 | Improved |
| increase alternating training iterations from 1 to 3 (ablation) | C2S-Net + CDC + CTS · ECSSD | weighted F-measure: 0.778 -> 0.837 | Improved |
| increase alternating training iterations from 3 to 5 (ablation) | C2S-Net + CDC + CTS · ECSSD | weighted F-measure: 0.837 -> 0.838 | No clear effect |
| increase unlabeled training data from 10K to 30K images (ablation) | C2S-Net + CDC + CTS + AT(3) · ECSSD | weighted F-measure: 0.837 -> 0.852 | Improved |
| increase unlabeled training data from 10K to 30K images (ablation) | C2S-Net + CDC + CTS + AT(3) · ECSSD | MAE: 0.059 -> 0.054 | Improved |
| use unlabeled 10K training instead of weakly supervised WSS | WSS · ECSSD | MAE: 0.103 -> 0.059 | Improved |
| use 30K unlabeled images instead of fully supervised Amulet | Amulet · ECSSD | F-measure: 0.915 -> 0.910 | Hurt |
| use 30K unlabeled images instead of fully supervised DSS | DSS · DUTS-TE | F-measure: 0.813 -> 0.807 | 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.
- Amulet — Amulet: Aggregating Multi-Level Convolutional Features for Salient Object Detection (ICCV 2017) · better on 2 of 3 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
- A Mutual Learning Method for Salient Object Detection With Intertwined Multi-Supervision (CVPR 2019 · similarity 0.92)
- Instance-Level Salient Object Segmentation (CVPR 2017 · similarity 0.90)
- Deep Embedding Features for Salient Object Detection (AAAI 2019 · similarity 0.90)
- Salient Object Detection With Pyramid Attention and Salient Edges (CVPR 2019 · similarity 0.90)
- BASNet: Boundary-Aware Salient Object Detection (CVPR 2019 · similarity 0.90)
- Salient Object Detection with Recurrent Fully Convolutional Networks (TPAMI 2018 · 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.
