Research map / Detection and segmentation / CNN features
Structural and Statistical Texture Knowledge Distillation and Learning for Segmentation
Deyi Ji, Feng Zhao, Hongtao Lu, Feng He Wu, Jieping Ye
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 adversarial learning to output-probability response KD (ablation) | response KD · Cityscapes val | mIoU: 70.51 -> 72.47 | Improved |
| add structural texture KD to response KD (ablation) | SSTKD student R18 · Cityscapes val | mIoU: 72.47 -> 74.80 | Improved |
| add statistical texture KD to response KD (ablation) | SSTKD student R18 · Cityscapes val | mIoU: 72.47 -> 75.89 | Improved |
| add both textures (ablation) | SSTKD student R18 · Cityscapes val | mIoU: 72.47 -> 76.65 | Improved |
| contourlet instead of FFT for structural texture (ablation) | structural KD · Cityscapes val | mIoU: 72.78 -> 74.80 | Improved |
| contourlet instead of wavelet (ablation) | structural KD · Cityscapes val | mIoU: 73.13 -> 74.80 | Improved |
| TIEM instead of GLCM for statistical texture (ablation) | statistical KD · Cityscapes val | mIoU: 72.97 -> 75.89 | Improved |
| add denoising step (ablation) | statistical texture KD · Cityscapes val | mIoU: 75.06 -> 75.89 | Improved |
| add adaptive sampling (ablation) | statistical texture KD · Cityscapes val | mIoU: 75.37 -> 75.89 | Improved |
| mild uncertainty-biased sampling (k=2, beta=0.7) vs k=1 beta=0 (ablation) | adaptive sampler · Cityscapes val | mIoU: 73.70 -> 75.89 | Improved |
| heavier biased sampling (k=8, beta=1.0) vs k=2 beta=0.7 (ablation) | adaptive sampler · Cityscapes val | mIoU: 75.89 -> 75.26 | Hurt |
| add SSTKD over CWD distillation | PSPNet-R18* · Cityscapes val | mIoU: 75.90 -> 76.65 | Improved |
Closest papers by meaning
- Structural and Statistical Texture Knowledge Distillation for Semantic Segmentation (CVPR 2022 · similarity 0.97)
- Learning Statistical Texture for Semantic Segmentation (CVPR 2021 · similarity 0.90)
- Deep Structure-Revealed Network for Texture Recognition (CVPR 2020 · similarity 0.86)
- Complex Wavelet Mutual Information Loss: A Multi-Scale Loss Function for Semantic Segmentation (ICML 2025 · similarity 0.85)
- Integrating Low-Level Visual Cues for Enhanced Unsupervised Semantic Segmentation (AAAI 2025 · similarity 0.85)
- Distilling Knowledge from Heterogeneous Architectures for Semantic Segmentation (AAAI 2025 · 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.
