Research map / Detection and segmentation / CNN features
Richer Convolutional Features for Edge Detection
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 |
|---|---|---|---|
| Use features from all convolutional layers rather than only the last layer of each stage | VGG16 edge detector · BSDS500 | ODS F-measure: .788 -> 0.806 | Helped |
| Use multiscale testing with scales 0.5, 1.0, and 1.5 (ablation) | RCF · BSDS500 | ODS F-measure: 0.806 -> 0.811 | Helped |
| Use multiscale testing (ablation) | RCF · BSDS500 | FPS: 30 FPS to 8 FPS | Hurt |
| Combine RGB and HHA predictions (ablation) | RCF · NYUD | ODS F-measure: .729 -> .757 | Helped |
| Use RCF-style features only in stages 1 and 2, instead of all stages (ablation) | RCF · BSDS500 | ODS F-measure: .806 -> .792 | Hurt |
| Use RCF-style features only in stages 2 and 4, instead of all stages (ablation) | RCF · BSDS500 | ODS F-measure: .798 -> .795 | Hurt |
| Use RCF-style features only in stages 4 and 5, instead of all stages (ablation) | RCF · BSDS500 | ODS F-measure: .798 -> .790 | Hurt |
| Use RCF-style features only in stages 1, 3, and 5, instead of all stages (ablation) | RCF · BSDS500 | ODS F-measure: .798 -> .794 | Hurt |
| Use RCF-style features only in stages 3, 4, and 5, instead of all stages (ablation) | RCF · BSDS500 | ODS F-measure: .798 -> .796 | Hurt |
| Replace all-layer RCF side outputs with HED side outputs (ablation) | RCF · BSDS500 | ODS F-measure: .798 -> .788 | Hurt |
| Add ReLU after the 1×1 convolution layers (ablation) | RCF · BSDS500 | ODS F-measure | Hurt |
| Use multiscale testing (ablation) | RCF · Multicue-Boundary | ODS F-measure: .817 (.004) -> .825 (.008) | Helped |
Closest papers by meaning
- Feature Pyramid Networks for Object Detection (CVPR 2017 · similarity 0.86)
- Deep Crisp Boundaries (CVPR 2017 · similarity 0.86)
- Learning Relaxed Deep Supervision for Better Edge Detection (CVPR 2016 · similarity 0.86)
- EDTER: Edge Detection With Transformer (CVPR 2022 · similarity 0.86)
- Harnessing Edge Information for Improved Robustness in Vision Transformers (AAAI 2024 · similarity 0.85)
- Deep Edge Filter: Return of the Human-Crafted Layer in Deep Learning (NeurIPS 2025 · 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.