Research map / Detection and segmentation / CNN features
RefineFace: Refinement Neural Network for High Performance Face Detection
Shifeng Zhang, Cheng Chi, Zhen Lei, Stan Ziqing Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 selective two-step regression (ablation) | RetinaNet baseline · WIDER FACE val | AP (Hard): 88.0 -> 88.8 | Improved |
| add selective two-step classification (ablation) | RetinaNet baseline · WIDER FACE val | AP (Hard): 88.0 -> 89.4 | Improved |
| add receptive field enhancement (ablation) | RetinaNet baseline · WIDER FACE val | AP (Hard): 88.0 -> 88.3 | Improved |
| add scale-aware margin loss (ablation) | RetinaNet baseline · WIDER FACE val | AP (Hard): 88.0 -> 89.1 | Improved |
| add feature supervision module (ablation) | RetinaNet baseline · WIDER FACE val | AP (Hard): 88.0 -> 88.7 | Improved |
| add RFE after STR and STC (ablation) | RefineFace · WIDER FACE val | AP (Hard): 90.1 -> 90.2 | Improved |
| combine all five modules versus baseline (ablation) | RetinaNet baseline · WIDER FACE val | AP (Hard): 88.0 -> 91.1 | Improved |
| apply two-step regression to pyramid level P2 (ablation) | RetinaNet baseline · WIDER FACE val | AP (Hard): 88.0 -> 87.5 | Hurt |
| apply two-step regression to high pyramid level P7 (ablation) | RetinaNet baseline · WIDER FACE val | AP (Hard): 88.0 -> 88.4 | Improved |
| apply two-step classification to high levels P5-P7 (ablation) | RetinaNet baseline · WIDER FACE val | AP (Hard): 88.0 -> 87.7 | No clear effect |
| add STC to reduce false positives (ablation) | RetinaNet · WIDER FACE val | false positives at 95% recall: 466,534 -> 103,586 | Improved |
| add STR at IoU threshold 0.8 (ablation) | RetinaNet · WIDER FACE val Hard | AP at IoU 0.8: 28.5 -> 38.2 | Improved |
Closest papers by meaning
- Selective Refinement Network for High Performance Face Detection (AAAI 2019 · similarity 0.94)
- Dynamic Refinement Network for Oriented and Densely Packed Object Detection (CVPR 2020 · similarity 0.87)
- Enriched Feature Guided Refinement Network for Object Detection (ICCV 2019 · similarity 0.86)
- Single-Shot Refinement Neural Network for Object Detection (CVPR 2018 · similarity 0.86)
- ProgressFace: Scale-Aware Progressive Learning for Face Detection (ECCV 2020 · similarity 0.86)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 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.
