A novel YOLO V5 framework for infrared image recognition
Yuquan Zhou, Zhiqiang Wang, Guodong Xie, Junrong Liao, Wenhao Huang, Chuwei Xiao · 2022 IEEE 5th International Conference on Automation, Electronics and Electrical Engineering (AUTEEE) · 2022
When it comes to face detection, problems of different illumination angles and uneven illumination distribution will inevitably produce some influences on the detection of human faces. In light of mentioned situation, this paper put forward an new framework of infrared face recognition algorithm based on YOLOv5 framework. Because of the small scale of indoor face dataset, the post-training model often overfits the brochure from the training set, leading to reduce generalization capability and other problems. Here, homomorphic filtering technology is used to increase the limited data sets. Then, the infrared face recognition model is trained based on YOLOv5. Experimental results indicate that the detection model achieves an accuracy of 93.55% and an F1-Sore of 87.37, which can adapt to detection tasks under different lighting changes and different shooting angles.