Research map / Detection and segmentation / CNN features
Intra-class Feature Variation Distillation for Semantic Segmentation
Yukang Wang, Wei Zhou, Tao Jiang, Xiang Bai, Yongchao Xu
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 KD loss to ResNet18 student (ablation) | PSPNet-ResNet18 · Cityscapes val | mIoU: 69.10 -> 70.51 (+1.41) | Improved |
| add adversarial loss to KD (ablation) | PSPNet-ResNet18 · Cityscapes val | mIoU: 70.51 -> 72.47 | Improved |
| add intra-class feature variation (IFV) loss to KD + adversarial (ablation) | PSPNet-ResNet18 · Cityscapes val | mIoU: 72.47 -> 74.54 | Improved |
| IFVD instead of SKD | PSPNet-ResNet18 · Cityscapes val | mIoU: 72.70 -> 74.54 | Improved |
| IFVD instead of SKD | PSPNet-ResNet18 (0.5) · Cityscapes val | mIoU: 61.60 -> 63.35 | Improved |
| IFVD instead of SKD | PSPNet-EfficientNet-B1 · Cityscapes val | mIoU: 63.13 -> 66.50 | Improved |
| IFVD instead of SKD | PSPNet-EfficientNet-B0 · Cityscapes test | mIoU: 61.80 -> 62.52 | Improved |
| IFVD instead of SKD | PSPNet-ResNet18 · CamVid test | mIoU: 71.0 -> 71.8 | Improved |
| IFVD instead of SKD | PSPNet-EfficientNet-B0 · CamVid test | mIoU: 63.9 -> 64.4 | Improved |
| IFVD instead of SKD | PSPNet-ResNet18 · Pascal VOC 2012 val | mIoU: 73.05 -> 74.05 | Improved |
| IFVD instead of SKD | PSPNet-EfficientNet-B0 · Pascal VOC 2012 val | mIoU: 70.24 -> 71.07 | Improved |
| use a global dataset-wide prototype instead of per-image class prototype (ablation) | IFVD · Cityscapes val | mIoU: 74.54 -> 73.86 (-0.68) | Hurt |
Closest papers by meaning
- Decomposed Knowledge Distillation for Class-Incremental Semantic Segmentation (NeurIPS 2022 · similarity 0.90)
- Class Similarity Weighted Knowledge Distillation for Continual Semantic Segmentation (CVPR 2022 · similarity 0.89)
- Structured Knowledge Distillation for Semantic Segmentation (CVPR 2019 · similarity 0.89)
- Distilling Knowledge from Heterogeneous Architectures for Semantic Segmentation (AAAI 2025 · similarity 0.88)
- $\mathbf{D^{3}S^{2}}$: Diffusion-Guided Dataset Distillation for Semantic Segmentation (NeurIPS 2026 · similarity 0.87)
- Wasserstein Distance Rivals Kullback-Leibler Divergence for Knowledge Distillation (NeurIPS 2024 · similarity 0.87)
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.
