Automated Deep Learning Framework for Accurate Human Identity and Gender Detection

Abstract

Future commercial uses for automated human identity verification include monitoring, surveillance, consumer profiling, and human-computer interaction. Using deep learning methods, this research examines the challenge of establishing a person's identity and gender from a single photograph of their hand. To begin, the hand pictures have been improved by pre-processing methods. Once the picture has been analysed, a convolutional neural network is used to determine which species of geese it is. To improve its ability to recognize individuals of a given gender, a network has been trained using only images of that gender. The framework has been designed using a variety of optimizers and k-fold cross-validation to increase the likelihood of success. The results of the experiment demonstrate that the objectives of individual identification and gender categorization were successfully met. Classifying persons based on their gender with a dorsal hand had an average accuracy of 97.75%, but with a palm hand, the accuracy drops to 96.79%. Finally, the proposed method outperforms previous methods in terms of accurately identifying and classifying individuals based on their gender.