Research map / Label-efficient and robust learning / Knowledge distillation
Inter-Region Affinity Distillation for Road Marking 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 inter-region affinity distillation to ERFNet using ResNet-101 as teacher | ERFNet · ApolloScape-test | mIoU: 40.4 -> 43.2 | Helped |
| Add inter-region affinity distillation to ERFNet using ResNet-101 as teacher | ERFNet · CULane-test | F1: 70.2 -> 72.4 | Helped |
| Add inter-region affinity distillation to ERFNet using ResNet-101 as teacher | ERFNet · LLAMAS-test | mAP: 0.570 -> 0.598 | Helped |
| Use all three moments (mean, variance, and skewness) in affinity distillation (ablation) | IntRA-KD · ApolloScape | mIoU | Helped |
| Add attention-map distillation alongside inter-region affinity distillation (ablation) | IntRA-KD · ApolloScape | mIoU: 43.2 mIoU | Helped |
| Distill high-level features rather than middle-level features (ablation) | IntRA-KD · ApolloScape | mIoU | Helped |
| Vary the loss coefficients among 0.05, 0.10, and 0.15 (ablation) | ERFNet-IntRA-KD · ApolloScape | mIoU: 43.18, 43.20, 43.17 | No clear effect |
| Vary the loss coefficients among 0.05, 0.10, and 0.15 (ablation) | ERFNet-IntRA-KD · CULane | F1: 72.36, 72.39, 72.38 | No clear effect |
| Vary the loss coefficients among 0.05, 0.10, and 0.15 (ablation) | ERFNet-IntRA-KD · LLAMAS | mAP: 0.597, 0.598, 0.598 | No clear effect |
| Use smoothed label regions as AOIs instead of raw ground-truth marking areas (ablation) | IntRA-KD · Road marking segmentation | contextual knowledge transfer | Helped |
Closest papers by meaning
- Structured Knowledge Distillation for Semantic Segmentation (CVPR 2019 · similarity 0.87)
- Distilling Knowledge from Heterogeneous Architectures for Semantic Segmentation (AAAI 2025 · similarity 0.85)
- RoadSceneBench: A Lightweight Benchmark for Mid-Level Road Scene Understanding (CVPR 2026 · similarity 0.85)
- Wasserstein Distance Rivals Kullback-Leibler Divergence for Knowledge Distillation (NeurIPS 2024 · similarity 0.84)
- DCSF-KD: Dynamic Channel-wise Spatial Feature Knowledge Distillation for Object Detection (AAAI 2025 · similarity 0.84)
- FreeKD: Knowledge Distillation via Semantic Frequency Prompt (CVPR 2024 · similarity 0.84)
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.