Research map / Detection and segmentation / Weakly supervised detection and segmentation
Joint Learning of Saliency Detection and Weakly Supervised Semantic Segmentation
Yu Zeng, Yunzhi Zhuge, Huchuan Lu, Lihe Zhang
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 saliency detection data and loss (multi-task) to stage-1 segmentation training (ablation) | SSNet-1 · PASCAL VOC 2012 val | mIoU: 33.1 -> 57.1 | Improved |
| use multi-task stage-1 predictions as pseudo labels for stage 2 instead of single-task ones (ablation) | SSNet-2 (single-task) · PASCAL VOC 2012 val | mIoU: 47.1 -> 62.7 | Improved |
| add saliency data and loss in stage 2 (multi-task) on top of multi-task stage 1 (ablation) | SSNet-2 · PASCAL VOC 2012 val | mIoU: 62.7 -> 63.3 | Improved |
| add multi-task training to saliency detection (ablation) | SSNet-2 · ECSSD | max F-measure: 0.899 -> 0.912 | Improved |
| add multi-task training to saliency detection (ablation) | SSNet-2 · ECSSD | MAE: 0.046 -> 0.047 | No clear effect |
| add CRF post-processing to saliency maps (ablation) | SSNet-MM · ECSSD | max F-measure: 0.912 -> 0.919 | Improved |
| use DenseNet-169 instead of VGG16 feature extractor | SSNet · PASCAL VOC 2012 val | mIoU: 57.1 -> 63.3 | Improved |
| use joint SSNet instead of MDC with pre-trained saliency model | weakly supervised segmentation · PASCAL VOC 2012 test | mIoU: 60.8 -> 64.3 | Improved |
| use joint SSNet instead of PAGR for saliency | saliency detection · ECSSD | max F-measure: 0.927 -> 0.919 | Hurt |
| use joint SSNet instead of RAS for saliency | saliency detection · ECSSD | MAE: 0.056 -> 0.045 | Improved |
Closest papers by meaning
- Learning to Detect Salient Object with Multi-source Weak Supervision (TPAMI 2021 · similarity 0.90)
- Multi-Source Weak Supervision for Saliency Detection (CVPR 2019 · similarity 0.90)
- Learning Saliency-Free Model with Generic Features for Weakly-Supervised Semantic Segmentation (AAAI 2020 · similarity 0.89)
- Railroad Is Not a Train: Saliency As Pseudo-Pixel Supervision for Weakly Supervised Semantic Segmentation (CVPR 2021 · similarity 0.89)
- Salient Object Detection with Recurrent Fully Convolutional Networks (TPAMI 2018 · similarity 0.89)
- STC: A Simple to Complex Framework for Weakly-Supervised Semantic Segmentation (TPAMI 2016 · 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.
