Research map / Detection and segmentation / CNN features
Learning Context-Aware Classifier for Semantic Segmentation
Read the paperCheck 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 a context-aware classifier | UperNet + Swin-Large · ADE20K val, single-scale | mIoU: 52.00 -> 52.87 | Helped |
| Add a context-aware classifier | UperNet + Swin-Large · COCO-Stuff 164K val, multi-scale | mIoU: 48.93 -> 50.00 | Helped |
| Add KL distillation to the full loss (ablation) | CAC · ADE20K | mIoU: 45.74 -> 46.91 | Helped |
| Replace vanilla KL with class-wise entropy KL (ablation) | CAC · ADE20K | mIoU: 45.72 -> 46.91 | Helped |
| Use an entropy mask from original predictions instead of oracle predictions (ablation) | CAC with class-wise entropy KL · ADE20K | mIoU: 46.91 -> 46.22 | Hurt |
| Use an entropy mask from estimated context-aware predictions instead of oracle predictions (ablation) | CAC with class-wise entropy KL · ADE20K | mIoU: 46.91 -> 46.58 | Hurt |
| Apply cosine similarity to the original classifier instead of dot product (ablation) | CAC pipeline · ADE20K | mIoU: 44.51 -> 43.89 | Hurt |
| Use dot product instead of cosine similarity for the context-aware classifier (ablation) | CAC · ADE20K | mIoU: 46.91 -> 45.42 | Hurt |
| Set the cosine scaling factor tau to 5 rather than 15 (ablation) | CAC · ADE20K | mIoU: 46.91 -> 45.39 | Hurt |
| Use projected prototypes without concatenating original classifier weights (ablation) | CAC · ADE20K | mIoU: 46.91 -> 46.21 | Hurt |
| Add projected prototypes and original classifier weights instead of concatenating them (ablation) | CAC · ADE20K | mIoU: 46.91 -> 46.53 | Hurt |
| Add original classifier weights as a residual after the projector (ablation) | CAC · ADE20K | mIoU: 46.91 -> 45.94 | Hurt |
Closest papers by meaning
- Context-Reinforced Semantic Segmentation (CVPR 2019 · similarity 0.88)
- Context Encoding for Semantic Segmentation (CVPR 2018 · similarity 0.87)
- Test-Time Adaptation of Vision-Language Models for Open-Vocabulary Semantic Segmentation (NeurIPS 2025 · similarity 0.86)
- Hunting Attributes: Context Prototype-Aware Learning for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.86)
- Distilling Spectral Graph for Object-Context Aware Open-Vocabulary Semantic Segmentation (CVPR 2025 · similarity 0.85)
- Unveiling the Knowledge of CLIP for Training-Free Open-Vocabulary 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.