Improving Gender Classification Using Huge Image Database with Deep Learning-Based Model
Abdul Khaliq · Zenodo (CERN European Organization for Nuclear Research) · 2023
Human gender classification received huge attention because of its diverse applications in various domains such as human-computer interaction (HCI) and computer-aided technology. It has previously been studied using various visual and machine learning algorithms but existing methods still face challenges for varying ages and ethnic groups. In order to meet this challenge, we take a step to improve our proposed model performance using a huge custom-built clean image database. In this paper, we recommend using a custom huge human face database to train the neural network-based model that can improve the gender classification accuracy of diverse ethnic groups. Thus, we prepared a huge gender image database of 200 thousand of images to enhance the performance of the proposed gender classification model. We cleaned and preprocessed the dataset to improve and speed up the model training procedure. The experimental outcome shows the highest accuracy of 95.7% with 98% confidence in the gender classification of an individual image.