Advanced Gender Detection Using Deep Learning Algorithms Through Hand X-Ray Images

Identifying the gender, race, age, and stature of the target during the forensic inquiry is a critical stage in various events such as accidents, bombings, terrorism, wars, and disasters. In this paper, an application has been developed that uses hand X-rays to identify and determine gender for medical applications such as special cases where diagnosing the gender is difficult, like accidents in which the hand is amputated and unknown, severe burns, and in old skeletal structures using deep learning models. For comparative purposes, GoogLeNet and ResNet-18 were employed. Gender determination using hand X-rays yielded positive results. The accuracy of gender detection in the model GoogLeNet (validation, training, test, and total) is (76.67%, 96.68%, 53.33%, and 89.5%) respectively, while the accuracy of gender detection in the model ResNet-18 (validation, training, test, and total) are (80%, 99.29%, 87.5%, 94.63%) respectively. The ResNet-18 model was adopted as the best model for gender detection and determination because high results were obtained. Simulation results showed acceptable results with high accuracy in diagnosis, where the highest gender determination rate was obtained through hand X-ray analysis at 94.63%.