Research map / Adversarial robustness and security / Attacks on text and graphs
SegMix: Co-occurrence Driven Mixup for Semantic Segmentation and Adversarial Robustness
Md Amirul Islam, Matthew D. Kowal, Konstantinos G. Derpanis, Neil D. B. Bruce
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 feature denoising stage (ablation) | Binding-CC (ResNet101) · PASCAL VOC 2012 val | mIoU: 76.4 -> 77.9 | Improved |
| add feature denoising stage (ablation) | Binding-CM (ResNet101) · PASCAL VOC 2012 val | mIoU: 78.3 -> 78.9 | Improved |
| add feature denoising stage (ablation) | Binding-CC (ResNet50) · PASCAL VOC 2012 val | mIoU: 75.4 -> 75.7 | Improved |
| add feature binding head (ablation) | Binding-CC (ResNet101) · PASCAL VOC 2012 val | mIoU: 76.1 -> 76.4 | Improved |
| add feature binding head (ablation) | Binding-CM (ResNet101) · PASCAL VOC 2012 val | mIoU: 77.9 -> 78.3 | Improved |
| add co-occurrence based blending to Mixup (ablation) | DeepLabv3-ResNet50 + Mixup · PASCAL VOC 2012 val | mIoU: 73.6 -> 74.2 | Improved |
| add co-occurrence based blending to Mixup (ablation) | DeepLabv3-ResNet101 + Mixup · PASCAL VOC 2012 val | mIoU: 76.2 -> 77.2 | Improved |
| add feature binding training on top of Mixup + co-occurrence (ablation) | DeepLabv3-ResNet50 · PASCAL VOC 2012 val | mIoU: 74.2 -> 76.1 | Improved |
| restrict unique categories and fix mixing ratio gamma to 0.9 above threshold (ablation) | Binding-CM (ResNet101) · PASCAL VOC 2012 val | mIoU: 77.0 -> 78.3 | Improved |
| apply Mixup to DeepLabv3 (ResNet50) | DeepLabv3-ResNet50 · PASCAL VOC 2012 val | mIoU: 75.1 -> 73.6 | Hurt |
| add Binding-CM training | DeepLabv3-ResNet101 · PASCAL VOC 2012 val | mIoU: 77.1 -> 78.9 | Improved |
| add Binding-CC training | DeepLabv3-ResNet101 · PASCAL VOC 2012 val, UAP ResNet152 attack | mIoU: 59.1 -> 69.1 | Improved |
Closest papers by meaning
- Adversarial Examples for Semantic Segmentation and Object Detection (ICCV 2017 · similarity 0.86)
- Rethinking Adversary in Semantic Segmentation: An Out-of-Distribution Perspective (ECCV 2026 · similarity 0.86)
- Separate and Conquer: Decoupling Co-occurrence via Decomposition and Representation for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.85)
- Increasing the Robustness of Semantic Segmentation Models with Painting-by-Numbers (ECCV 2020 · similarity 0.85)
- On the Robustness of Semantic Segmentation Models to Adversarial Attacks (CVPR 2018 · similarity 0.85)
- MultiSeg: Semantically Meaningful, Scale-Diverse Segmentations From Minimal User Input (ICCV 2019 · similarity 0.85)
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.
