Research map / Detection and segmentation / Weakly supervised detection and segmentation
ECS-Net: Improving Weakly Supervised Semantic Segmentation by Using Connections Between Class Activation Maps
Kunyang Sun, Haoqing Shi, Zhengming Zhang, Yongming Huang
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 erasing of discriminative regions without connections between CAMs (ablation) | CAM baseline · VOC 2012 train | mIoU: 47.4 -> 48.5 | Improved |
| add segment supervision from connections between original and erased CAMs (CE loss) (ablation) | ECS-Net · VOC 2012 train | mIoU: 48.5 -> 55.1 | Improved |
| add other improvements (0.5 0.8 model, multi dilated overlay module) (ablation) | ECS-Net · VOC 2012 train | mIoU: 55.1 -> 56.6 | Improved |
| add CRF post-processing (ablation) | ECS-Net · VOC 2012 train | mIoU: 56.6 -> 58.6 | Improved |
| use rescale factor 0.5 instead of 1 for first input image (ablation) | ECS-Net · VOC 2012 train | mIoU: 55.1 -> 56.1 | Improved |
| use rescale factor 0.7 instead of 1 for first input image (ablation) | ECS-Net · VOC 2012 train | mIoU: 55.1 -> 55.7 | Improved |
| raise threshold theta from 0.6 to 0.8 (ablation) | ECS-Net · VOC 2012 train | mIoU: 53.8 -> 55.1 | Improved |
| raise threshold theta from 0.8 to 0.9 (ablation) | ECS-Net · VOC 2012 train | mIoU: 55.1 -> 54.9 | Hurt |
| use CAM heatmap H' rather than refine heatmap E' to produce segment labels (ablation) | ECS-Net · VOC 2012 train | mIoU: 54.2 -> 55.1 | Improved |
| use ECS-Net CAMs instead of SEAM-refined CAMs | CAM · VOC 2012 train | mIoU: 55.4 -> 56.6 | Improved |
| use ECS-Net pseudo labels with ResNet-38 instead of SEAM | DeepLabv1 · VOC 2012 val | mIoU: 64.5 -> 66.6 | Improved |
Later papers that compared with this method
1 later accepted paper reports ECS-Net as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Adaptive Spatial-BCE Loss for Weakly Supervised Semantic Segmentation (ECCV 2022 · better on 1 of 1 shared result)
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.
- FickleNet — FickleNet: Weakly and Semi-Supervised Semantic Image Segmentation Using Stochastic Inference (CVPR 2019) · better on 1 of 1 shared result
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
- Unlocking the Potential of Ordinary Classifier: Class-Specific Adversarial Erasing Framework for Weakly Supervised Semantic Segmentation (ICCV 2021 · similarity 0.92)
- Complementary Patch for Weakly Supervised Semantic Segmentation (ICCV 2021 · similarity 0.91)
- Learning Integral Objects With Intra-Class Discriminator for Weakly-Supervised Semantic Segmentation (CVPR 2020 · similarity 0.91)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 2024 · similarity 0.89)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.89)
- Separate and Conquer: Decoupling Co-occurrence via Decomposition and Representation for Weakly Supervised Semantic Segmentation (CVPR 2024 · 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.
