Progressive Calibration Networks (PCN) is an accurate rotation-invariant face detector running at real-time speed on CPU, published in CVPR 2018. This is a binary library for PCN. In this implementation, we don't use network quantization or compression, and the program runs on CPU with a single thread.
Some detection results can be viewed in the following illustrations:
PCN is designed aiming for low time-cost. We compare PCN's speed with other rotation-invariant face detectors' on standard VGA images(640x480) with 40x40 minimum face size. The detectors run on a desktop computer with 3.4GHz CPU, GTX Titan X. The speed results together with the recall rate at 100 false positives on multi-oriented FDDB are shown in the following table. Detailed experiment settings can be found in our paper.
Set minimum size of faces to detect (size >= 20)
detector.SetMinFaceSize(size);
Set scaling factor of image pyramid (1.4 <= factor <= 1.6)
detector.SetImagePyramidScaleFactor(factor);
Set score threshold of detected faces (0 <= thresh <= 1)
detector.SetScoreThresh(thresh);
Smooth the face boxes or not (smooth = true or false, recommend using it on video to get stabler face boxes)
detector.SetVideoSmooth(smooth);
See picture.cpp and video.cpp for details. If you want to reproduce the results on FDDB, set size and factor as 20 and 1.414 respectively, or you can adjust them according to your application.
- Linux
- Caffe (recommend using OpenBLAS)
- OpenCV (2.4.10, or other compatible version)
This code is distributed under the BSD 2-Clause license.
If you find PCN useful in your research, please consider citing:
@inproceedings{shiCVPR18pcn,
Author = {Xuepeng Shi and Shiguang Shan and Meina Kan and Shuzhe Wu and Xilin Chen},
Title = {Real-Time Rotation-Invariant Face Detection with Progressive Calibration Networks},
Booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
Year = {2018}
}
Xuepeng Shi, xuepeng.shi@vipl.ict.ac.cn