Research map / Detection and segmentation / CNN features
Semantic Image Segmentation With Task-Specific Edge Detection Using CNNs and a Discriminatively Trained Domain Transform
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 |
|---|---|---|---|
| Replace image gradients with jointly learned, task-specific EdgeNet edges for domain-transform filtering | DeepLab segmentation on PASCAL VOC 2012 · PASCAL VOC 2012 val | mIOU (%): 62.25 -> 66.35 | Helped |
| Use EdgeNet edges instead of Structured Edges | Domain-transform filtering · PASCAL VOC 2012 val | mIOU (%): 64.89 -> 66.35 | Helped |
| Apply dense CRF after DT-EdgeNet filtering | DT-EdgeNet · PASCAL VOC 2012 val | mIOU (%): 66.35 -> 68.44 | Helped |
| Use DT-EdgeNet instead of dense CRF post-processing | DeepLab segmentation · PASCAL VOC 2012 test, ImageNet pretrained | mIOU (%): 70.3 -> 69.0 | Hurt |
| Use DT-EdgeNet instead of dense CRF post-processing | DeepLab segmentation · PASCAL VOC 2012 test, MS-COCO pretrained | mIOU (%): 72.7 -> 71.7 | Hurt |
| Apply domain-transform filtering to DeepLab-CRF-Attention | DeepLab-CRF-Attention · PASCAL VOC 2012 test, MS-COCO pretrained | mIOU (%): 75.7 -> 76.3 | Helped |
| Use conv3_3 features for EdgeNet rather than no EdgeNet feature fusion (ablation) | DeepLab with DT · PASCAL VOC 2012 val | mIOU (%): 62.25 -> 65.64 | Helped |
| Add conv2_2 features to conv3_3 (ablation) | EdgeNet · PASCAL VOC 2012 val | mIOU (%): 65.64 -> 65.75 | Helped |
| Add conv4_3 features to conv2_2 + conv3_3 (ablation) | EdgeNet · PASCAL VOC 2012 val | mIOU (%): 65.75 -> 66.03 | Helped |
| Add conv5_3 features to conv2_2 + conv3_3 + conv4_3 (ablation) | EdgeNet · PASCAL VOC 2012 val | mIOU (%): 66.03 -> 65.94 | Hurt |
| Add conv1_2 features to conv2_2 + conv3_3 + conv4_3 (ablation) | EdgeNet · PASCAL VOC 2012 val | mIOU (%): 66.03 -> 65.89 | Hurt |
| Use three DT iterations rather than fewer iterations (ablation) | DT-EdgeNet · PASCAL VOC 2012 val | mIOU | Helped |
Closest papers by meaning
- Semantic Segmentation With Boundary Neural Fields (CVPR 2016 · similarity 0.90)
- Efficient Piecewise Training of Deep Structured Models for Semantic Segmentation (CVPR 2016 · similarity 0.89)
- Revisiting Dilated Convolution: A Simple Approach for Weakly- and Semi-Supervised Semantic Segmentation (CVPR 2018 · similarity 0.89)
- Unsupervised Semantic Segmentation Through Depth-Guided Feature Correlation and Sampling (CVPR 2024 · similarity 0.87)
- Unlocking the Potential of Pre-trained Vision Transformers for Few-Shot Semantic Segmentation through Relationship Descriptors (CVPR 2024 · similarity 0.87)
- SemiDAViL: Semi-supervised Domain Adaptation with Vision-Language Guidance for Semantic Segmentation (CVPR 2025 · 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.