Research map / Detection and segmentation / CNN features
Learning Debiased and Disentangled Representations for Semantic Segmentation
Sanghyeok Chu, Dongwan Kim, Bohyung Han
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 DropClass training | HRNet-W18-v1 · Cityscapes | mIoU: 69.3 -> 70.7 | Improved |
| add DropClass training | HRNet-W18-v2 · Cityscapes | mIoU: 75.1 -> 76.6 | Improved |
| add DropClass training | DeepLabV3 · Cityscapes | mIoU: 67.2 -> 67.7 | Improved |
| add DropClass training | HRNet-W18-v2 · Cityscapes | mIoU of 9 rarest classes: 67.1 -> 69.5 | Improved |
| add DropClass training | DeepLabV3 · Cityscapes | mIoU of 9 rarest classes: 55.9 -> 57.0 | Improved |
| remove median frequency reweighting from both models | DropClass · Cityscapes | mIoU: 69.3 -> 70.1 | Improved |
| resample motorcycle images twice | Baseline · Cityscapes | motorcycle IoU: 39.9 -> 46.9 | Improved |
| resample motorcycle images twice | Baseline · Cityscapes | mIoU: 69.2 -> 69.1 | Hurt |
| DropClass instead of motorcycle resampling | Baseline + 2x motorcycle · Cityscapes | motorcycle IoU: 46.9 -> 49.7 | Improved |
| remove suppression loss L_sup (ablation) | DropClass · Cityscapes | mIoU: 70.7 -> 70.2 | Hurt |
| remove suppression loss L_sup (ablation) | DropClass · Cityscapes | mIoU of 9 rarest classes: 59.7 -> 58.6 | Hurt |
| drop random class labels with cross-entropy only (Ignore Label) (ablation) | Baseline · Cityscapes | mIoU: 69.3 -> 69.2 | No clear effect |
Closest papers by meaning
- Towards Unbiased Learning in Semi-Supervised Semantic Segmentation (ICLR 2025 · similarity 0.88)
- Learning Yourself: Class-Incremental Semantic Segmentation with Language-Inspired Bootstrapped Disentanglement (ICCV 2025 · similarity 0.87)
- ExFuse: Enhancing Feature Fusion for Semantic Segmentation (ECCV 2018 · similarity 0.87)
- Consistent-Separable Feature Representation for Semantic Segmentation (AAAI 2021 · similarity 0.87)
- Diffusion-Driven Two-Stage Active Learning for Low-Budget Semantic Segmentation (NeurIPS 2025 · similarity 0.86)
- Tripod: Three Complementary Inductive Biases for Disentangled Representation Learning (ICML 2024 · similarity 0.86)
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.
