Prediction of individuals age and gender based on fingerphoto

This thesis work involves a Deep Recurrent Learning (DRL) model called the Deep Recurrent Fingerphotos Network (DRFN), which performs individual recognition, age estimation, and gender prediction based on finger photo images. The architecture of the proposed DRFN model consists of an input layer, several hidden layers, an output layer, and a significant feedback connection. This connection allows the sequential processing of finger photo images taken from an individual's ten fingers, enabling the DRFN to dynamically adjust the importance of each finger photo. To test the accuracy of the proposed model, researchers have collected a dataset consisting entirely of original images, featuring samples from individuals. For each individual, there are photos of all ten fingers taken from ten different angles. The tasks performed within the scope of the study on this original dataset include individual recognition, gender determination, age estimation, and the evaluation of performance results. Additionally, the study includes extensive experiments and comparisons between the proposed DRFN and other well-known deep learning networks used for feature extraction such as AlexNet, LeNet, GoogleNet, VGG16, and ResNet-50. These evaluations also utilized various classifiers including Support Vector Machines (SVM), Random Forest, Multilayer Perceptron (MLP), Linear Regression, and Logistic Regression. A wide range of performance metrics such as Precision, Recall, and F1-Score were used in the evaluation of the analyses. The results demonstrate that the proposed DRFN model performs better than the compared algorithms and that finger photos can be used in fields such as individual recognition, gender recognition, and age estimation. This thesis not only proposes a DRL-based framework for individual verification using finger photos but also demonstrates its effectiveness through detailed experiments and comparative analyses.