Face Detection and Gender Classification by YOLO Algorithm

Gender classification is a fundamental computer vision problem used in a wide range of applications in surveillance and marketing. The work in this paper is to determine the gender classification ability of the You Only Look Once (YOLO) algorithm using deep learning. YOLO is one of the most accurate object detection models that can detect multiple objects in a video or picture in real-time. In this work, various versions of YOLO (YOLOv3 to YOLOv9) were compared to determine the most accurate and efficient model for gender classification. The work utilized a collection of 361 test images of male and female subjects in different scenarios of their settings, and the models' performance was gauged in terms of key metrics of Precision, Recall, and F1-score. The analysis of performance also confirmed that YOLOv9 was even better compared to its counterparts, registering a mean average precision (mAP) of 97%, a Precision of 86.8%, a Recall of 86.1%, and an F1-score of 86.54%. The processing time of the model was 0.332 seconds per picture, or a frame per second (FPS) of 3.00. The confusion matrix also recorded 157 true positives (TP), 25 false negatives (FN), 23 false positives (FP), and 156 true negatives (TN), a reflection of a highly balanced classification performance. The results confirm that YOLOv9 is highly accurate and efficient to use in gender classification in practical applications.