Research map / Detection and segmentation / CNN features
Knowledge Adaptation for Efficient Semantic Segmentation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 affinity distillation (ablation) | MobileNetV2-16s · Pascal VOC trainaug to val | mIOU(%): 70.57 to 71.53 | Helped |
| add knowledge adaptation to affinity distillation (ablation) | MobileNetV2-16s · Pascal VOC trainaug to val | mIOU(%): 71.53 to 72.50; another 0.97 point | Helped |
| use knowledge adaptation and affinity distillation instead of the student baseline | MobileNetV2 · Pascal VOC 2012 val | mIOU(%): 70.57 -> 72.50 | Helped |
| apply output-logit knowledge distillation at temperatures 2, 4, or 8 (ablation) | MobileNetV2 · Pascal VOC trainaug to val | mIOU(%): 71.32; 71.21; 70.74 | Mixed |
| match intermediate representations with FitNet (ablation) | MobileNetV2 · Pascal VOC trainaug to val | mIOU(%): 70.57 -> 71.30 | Helped |
| replace FitNet matching with the proposed translated-knowledge method (ablation) | MobileNetV2 · Pascal VOC trainaug to val | mIOU(%): 71.30 to 72.50; 1.2 points | Helped |
| distill from Xception-41 rather than ResNet-50 (ablation) | MobileNetV2-1.0 · Pascal VOC 2012 val | mIOU(%): 72.50 vs 72.40 | No clear effect |
| increase MobileNetV2 width multiplier to 1.3 under the proposed method (ablation) | MobileNetV2 · Pascal VOC 2012 val | mIOU(%): 72.50 to 74.26; 1.66 | Helped |
| increase MobileNetV2 width multiplier to 1.4 under the proposed method (ablation) | MobileNetV2 · Pascal VOC 2012 val | mIOU(%): 73.36 to 74.07; 0.71 | Helped |
| apply the proposed method | MobileNetV2 · Pascal Context val | mIOU(%): 39.9 to 41.2; 1.3 points | Helped |
| apply the proposed method | MobileNetV2 · Cityscapes test | mIOU (%): 70.2 to 72.7; 2.5 points | Helped |
| apply the proposed method | MobileNetV2 · Cityscapes val | mIOU (%): 68.9 to 71.0 | Helped |
Closest papers by meaning
- Structured Knowledge Distillation for Semantic Segmentation (CVPR 2019 · similarity 0.90)
- Augmentation-Free Dense Contrastive Knowledge Distillation for Efficient Semantic Segmentation (NeurIPS 2023 · similarity 0.90)
- Generalizable Knowledge Distillation from Vision Foundation Models for Semantic Segmentation (CVPR 2026 · similarity 0.89)
- Rethinking Semantic Segmentation From a Sequence-to-Sequence Perspective With Transformers (CVPR 2021 · similarity 0.88)
- 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)
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.