Research map / Detection and segmentation / CNN features
Selective Refinement Network for High Performance Face Detection
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 STC to RetinaNet (ablation) | RetinaNet · WIDER FACE validation Easy, Medium, Hard | AP: 95.1% to 95.3%; 93.9% to 94.4%; 88.0% to 89.4% | Helped |
| Add STC to RetinaNet (ablation) | RetinaNet · WIDER FACE validation | false positives across recall rates: 3 -> 1; 24 -> 20; 126 -> 101; 2801 -> 2124; 27644 -> 13163; 466534 -> 103586 | Helped |
| Add STR to RetinaNet (ablation) | RetinaNet · WIDER FACE validation Easy, Medium, Hard | AP: 0.8%, 0.9% and 0.8% AP improvements | Helped |
| Add STR to RetinaNet (ablation) | RetinaNet · WIDER FACE Hard | AP at IoU thresholds 0.5, 0.6, 0.7, 0.8: 88.1 -> 88.8; 76.4 -> 83.4; 57.8 -> 66.5; 28.5 -> 38.2 | Helped |
| Add two-step classification at P2, P3, and P4 (ablation) | RetinaNet · WIDER FACE validation Easy, Medium, Hard | AP: P2: 95.2, 94.2, 88.9; P3: 95.2, 94.3, 88.7; P4: 95.2, 94.1, 88.5; baseline: 95.1, 93.9, 88.0 | Helped |
| Add two-step classification at P5, P6, or P7 (ablation) | RetinaNet · WIDER FACE validation | AP: P5: 95.0, 93.9, 87.8; P6: 95.1, 93.7, 88.0; P7: 95.0, 93.9, 87.7; baseline: 95.1, 93.9, 88.0 | Hurt |
| Add two-step regression at P2, P3, or P4 (ablation) | RetinaNet · WIDER FACE validation | AP: P2: 94.8, 93.4, 87.5; P3: 94.3, 93.7, 87.7; P4: 94.8, 93.9, 87.0; baseline: 95.1, 93.9, 88.0 | Hurt |
| Add two-step regression at P5, P6, or P7 (ablation) | RetinaNet · WIDER FACE validation | AP: P5: 95.4, 94.2, 88.2; P6: 95.7, 94.4, 88.2; P7: 95.6, 94.6, 88.4; baseline: 95.1, 93.9, 88.0 | Helped |
| Add RFE to STC+STR (ablation) | SRN · WIDER FACE validation Easy, Medium, Hard | AP: 0.3%, 0.3%, and 0.1% | Helped |
Closest papers by meaning
- Single-Shot Refinement Neural Network for Object Detection (CVPR 2018 · similarity 0.85)
- FA-RPN: Floating Region Proposals for Face Detection (CVPR 2019 · similarity 0.84)
- Dynamic Refinement Network for Oriented and Densely Packed Object Detection (CVPR 2020 · similarity 0.84)
- DSFD: Dual Shot Face Detector (CVPR 2019 · similarity 0.84)
- CRFace: Confidence Ranker for Model-Agnostic Face Detection Refinement (CVPR 2021 · similarity 0.83)
- BFBox: Searching Face-Appropriate Backbone and Feature Pyramid Network for Face Detector (CVPR 2020 · similarity 0.83)
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.