Optimization of Face Detection Algorithm based on MTCNN

Gaiping Liu ； Jianmei Xiao ； Xihuai Wang

《International Core Journal of Engineering》 7卷8期 (2021/08) Pp. 456-464

https://doi.org/10.6919/ICJE.202108_7(8).0067

摘要

Face detection is an important fundamental problem in computer vision, and it is a key step for future applications such as face analysis, face verification, face labeling and retrieval. With the development of image processing and deep learning, face detection and recognition have been widely used in all aspects of life, so face detection technology has put forward higher requirements. In practice, the image acquired in face detection is greatly affected by the environment, the image is fuzzy and contains noise, and there are some difficulties in detection. In this paper, based on the MTCNN face detection algorithm to solve the above problems, first of all, the image using SRGAN ultra-high resolution restoration technology to image clearness processing, make the face clearer, more convenient detection. In order to better process the feature map, the InceptionV2 module was introduced to optimize the network structure of MTCNN. It further improves the accuracy of face detection in complex background. The training was validated on WIDER FACE and CelebA datasets. The final accuracy of the optimized algorithm can reach 98.8%, 2.4% higher than that of the unoptimized network, which better meets the application requirements of modern society for face detection.

關鍵字

Face Detection ； Network Optimization ； Inception Module ； Image Sharpening

參考文獻

Shifeng Zhang, Cheng Chi, Zhen Lei, Stan Z. Li: “RefineFace: Refinement Neural Network for High Performance Face Detection”, 2019; [arXiv:1909.04376].

Xu Tang, Daniel K. Du, Zeqiang He, Jingtuo Liu: “PyramidBox: A Context-assisted Single Shot Face Detector”, 2018, ECCV2018; [arXiv:1803.07737].

Christian Ledig, Lucas Theis, Ferenc Huszar, Jose Caballero, Andrew Cunningham, Alejandro Acosta, Andrew Aitken, Alykhan Tejani, Johannes Totz, Zehan Wang, Wenzhe Shi: “Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network”, 2016; [arXiv:1609.04802].

ZHANG K P, ZHANG Z P, LI Z F, et al. Joint face detection and align⁃ment using multi-task cascaded convolutional networks [J]. IEEE Sig⁃nal Processing Letters, 2016.23 (10): 1499-1503.

Shifeng Zhang, Xiangyu Zhu, Zhen Lei, Hailin Shi, Xiaobo Wang, Stan Z. Li: “FaceBoxes: A CPU Real-time Face Detector with High Accuracy”, 2017; [arXiv:1708.05234].