Research map / Detection and segmentation / CNN features
A Fast and Accurate Unconstrained Face Detector
Shengcai Liao, Anil K. Jain, 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 |
|---|---|---|---|
| use NPD features instead of Haar features (same CART learner) (ablation) | CART-based face detector · FDDB (discrete metric, FP=1) | detection rate: about 6% | Improved |
| use NPD features instead of LBP features (same CART learner) (ablation) | CART-based face detector · FDDB (discrete metric, FP=1) | detection rate: about 19% | Improved |
| use NPD features instead of POF features (same CART learner) (ablation) | CART-based face detector · FDDB (discrete metric, FP=1) | detection rate: about 15% | Improved |
| use NPD features instead of Haar, LBP, POF (continuous metric) (ablation) | CART-based face detector · FDDB (continuous metric, FP=1) | detection rate: about 4%, 13%, and 10% | Improved |
| use NPD instead of the NPD2 variant sqrt-normalized feature (ablation) | CART-based face detector · FDDB | ROC curve | Improved |
| replace stump weak learners with CART trees (ablation) | NPD cascade detector · FDDB (discrete metric) | detection rate: about 0% - 17% | Improved |
| replace stump weak learners with CART trees (ablation) | NPD cascade detector · FDDB (continuous metric) | detection rate: about 0% - 11% | Improved |
| replace stump with CART trees (ablation) | NPD cascade detector · Training set of [13] | #trees: 1,597 -> 176 | Improved |
| replace CART with deep quadratic tree (DQT) (ablation) | NPD cascade detector · FDDB | average feature evaluations: 34.4 -> 23.5 | Improved |
| replace CART with deep quadratic tree (DQT) (ablation) | NPD cascade detector · FDDB | detection rate | Improved |
| use soft cascade instead of classic cascade with NPD+DQT (ablation) | NPD+DQT detector · FDDB | average feature evaluations: 23.5 -> 18.6 | Improved |
| use soft cascade instead of classic cascade (ablation) | NPD+DQT detector · FDDB | detection performance | No clear effect |
Closest papers by meaning
- MogFace: Towards a Deeper Appreciation on Face Detection (CVPR 2022 · similarity 0.86)
- Unconstrained Still/Video-Based Face Verification with Deep Convolutional Neural Networks (IJCV 2017 · similarity 0.86)
- Sample and Computation Redistribution for Efficient Face Detection (ICLR 2022 · similarity 0.85)
- Finding Tiny Faces (CVPR 2017 · similarity 0.84)
- Faceness-Net: Face Detection through Deep Facial Part Responses (TPAMI 2017 · similarity 0.84)
- ProgressFace: Scale-Aware Progressive Learning for Face Detection (ECCV 2020 · 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.
