Research map / Detection and segmentation / CNN features
Seeing Small Faces From Robust Anchor's Perspective
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 |
|---|---|---|---|
| Replace baseline feature-map generation with bilinear upsampling (ablation) | ResNet-101 anchor-based detector · Wider Face validation Hard | AP: 0.705 -> 0.710 | Helped |
| Enlarge the feature map with bilinear upsampling and a skip connection (ablation) | ResNet-101 anchor-based detector · Wider Face validation Hard | AP: 0.705 -> 0.778 | Helped |
| Use dilated convolution to enlarge the feature map (ablation) | ResNet-101 anchor-based detector · Wider Face validation Hard | AP: 0.705 -> 0.779 | Helped |
| Add one shifted 16×16 anchor (ablation) | +DC · Wider Face validation Hard | AP: 0.779 -> 0.781 | Helped |
| Add one shifted 16×16 anchor and one shifted 32×32 anchor (ablation) | +DC · Wider Face validation Hard | AP: 0.779 -> 0.782 | Helped |
| Add three shifted 16×16 anchors (ablation) | +DC · Wider Face validation Hard | AP: 0.779 -> 0.786 | Helped |
| Add three shifted 16×16 anchors and one shifted 32×32 anchor (ablation) | +DC · Wider Face validation Hard | AP: 0.779 -> 0.781 | Helped |
| Add three shifted anchors at both 16×16 and 32×32 scales (ablation) | +DC · Wider Face validation Hard | AP: 0.779 -> 0.779 | No clear effect |
| Add face-shift jittering during training (ablation) | +DC · Wider Face validation Hard | AP: 0.779 -> 0.788 | Helped |
| Add face-shift jittering during training (ablation) | +DC+16x3 · Wider Face validation Hard | AP: 0.786 -> 0.789 | Helped |
| Assign hard faces to the top 3 anchors (ablation) | +DC+16x3+SJ · Wider Face validation Hard | AP: 0.789 -> 0.793 | Helped |
| Assign hard faces to the top 5 anchors (ablation) | +DC+16x3+SJ · Wider Face validation Hard | AP: 0.789 -> 0.795 | Helped |
Closest papers by meaning
- AnchorFace: An Anchor-based Facial Landmark Detector Across Large Poses (AAAI 2021 · similarity 0.88)
- HAMBox: Delving Into Mining High-Quality Anchors on Face Detection (CVPR 2020 · similarity 0.88)
- Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection (CVPR 2020 · similarity 0.87)
- MogFace: Towards a Deeper Appreciation on Face Detection (CVPR 2022 · similarity 0.87)
- AnchorFace: Boosting TAR@FAR for Practical Face Recognition (AAAI 2022 · similarity 0.86)
- On the Use of Anchoring for Training Vision Models (NeurIPS 2024 · similarity 0.82)
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.