Research map / Detection and segmentation / CNN features
Context Encoding 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 Encoding Module (ablation) | FCN with ResNet-50 · PASCAL-Context (59 classes) | mIoU: 41.0 -> 47.6 | Helped |
| Add Semantic Encoding Loss (ablation) | EncNet with ResNet-50 · PASCAL-Context (59 classes) | mIoU: 47.6 -> 49.2 | Helped |
| Use ResNet-101 instead of ResNet-50 (ablation) | EncNet with SE-loss · PASCAL-Context (59 classes) | mIoU: 49.2 -> 51.7 | Helped |
| Add multi-size evaluation (ablation) | EncNet with ResNet-101 and SE-loss · PASCAL-Context (59 classes) | mIoU: 51.7 -> 52.6 | Helped |
| Set SE-loss weight to 0.2 (ablation) | EncNet · PASCAL-Context | mIoU | Helped |
| Increase codeword count, using K=32 rather than K=0 global average pooling (ablation) | Encoding Layer in EncNet · PASCAL-Context | mIoU | Helped |
| Replace the ResNet-50 FCN baseline with EncNet | ResNet-50 FCN · ADE20K validation | mIoU: 34.38 -> 41.11 | Helped |
| Use Context Encoding Modules instead of the baseline ResNet feature pathway | 14-layer ResNet 64d · CIFAR-10 test | Error rate: 4.93 -> 3.96 | Helped |
| Use EncNet rather than SE-Blocks | 14-layer ResNet with similar-scale 64d configuration · CIFAR-10 test | Error rate: 4.65 -> 3.96 | Helped |
| Increase EncNet configuration from 16k64d to 32k128d (ablation) | 14-layer EncNet · CIFAR-10 test | Error rate: 3.96 -> 3.45 | Helped |
Closest papers by meaning
- Rethinking Semantic Segmentation From a Sequence-to-Sequence Perspective With Transformers (CVPR 2021 · similarity 0.90)
- Context-Reinforced Semantic Segmentation (CVPR 2019 · similarity 0.89)
- Adaptive Pyramid Context Network for Semantic Segmentation (CVPR 2019 · similarity 0.89)
- Fully Convolutional Neural Networks with Full-Scale-Features for Semantic Segmentation (AAAI 2017 · similarity 0.89)
- Context Prior for Scene Segmentation (CVPR 2020 · similarity 0.89)
- Contextrast: Contextual Contrastive Learning for Semantic Segmentation (CVPR 2024 · similarity 0.88)
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.