Research map / Detection and segmentation / CNN features
S3FD: Single Shot Scale-Invariant Face Detector
Shifeng Zhang, Xiangyu Zhu, Zhen Lei, Hailin Shi, Xiaobo Wang, Stan Z. 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 |
|---|---|---|---|
| use scale-equitable framework (multi-layer anchors, scaled to stride) instead of RPN-face | anchor-based face detector · WIDER FACE val Hard | mAP (%): 73.7 -> 82.3 | Improved |
| use scale-equitable framework instead of SSD-face | anchor-based face detector · WIDER FACE val Hard | mAP (%): 71.6 -> 82.3 | Improved |
| use scale-equitable framework instead of SSD-face | anchor-based face detector · WIDER FACE val Easy | mAP (%): 92.1 -> 92.6 | Improved |
| add scale compensation anchor matching (ablation) | S3FD(F) · WIDER FACE val Easy | mAP (%): 92.6 -> 93.5 | Improved |
| add scale compensation anchor matching (ablation) | S3FD(F) · WIDER FACE val Medium | mAP (%): 91.6 -> 92.0 | Improved |
| add scale compensation anchor matching (ablation) | S3FD(F) · WIDER FACE val Hard | mAP (%): 82.3 -> 84.5 | Improved |
| add max-out background label (ablation) | S3FD(F+S) · WIDER FACE val Easy | mAP (%): 93.5 -> 93.7 | Improved |
| add max-out background label (ablation) | S3FD(F+S) · WIDER FACE val Medium | mAP (%): 92.0 -> 92.4 | Improved |
| add max-out background label (ablation) | S3FD(F+S) · WIDER FACE val Hard | mAP (%): 84.5 -> 85.2 | Improved |
| add Conv3_3 detection layer (smallest anchors) (ablation) | S3FD · WIDER FACE val Hard | mAP change when removed (%): -24.7 | Improved |
| add Conv3_3 detection layer (smallest anchors) (ablation) | S3FD · WIDER FACE val Easy and Medium | mAP change when removed (%): +0.3 Easy, +0.5 Medium | Hurt |
| add Conv5_3 detection layer (ablation) | S3FD · WIDER FACE val Easy | mAP change when removed (%): -25.8 | Improved |
Closest papers by meaning
- Group Sampling for Scale Invariant Face Detection (CVPR 2019 · similarity 0.88)
- Scale-Aware Face Detection (CVPR 2017 · similarity 0.88)
- Finding Tiny Faces (CVPR 2017 · similarity 0.87)
- Seeing Small Faces From Robust Anchor's Perspective (CVPR 2018 · similarity 0.87)
- DSFD: Dual Shot Face Detector (CVPR 2019 · similarity 0.87)
- ProgressFace: Scale-Aware Progressive Learning for Face Detection (ECCV 2020 · similarity 0.86)
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.
