Face-based age and gender classification using deep learning model

Abstract

The classification of the age and gender of face image data is a significant study subject with numerous applications. Convolution Neural Networks (CNNs) model is the effective technique to classify data, particularly for unrestricted real-world faces. In this current study, we introduce an innovative CNN-based method specifically designed for the extraction of distinguishing features from real-world facial data. These extracted features play a crucial role in the categorization and classification of images, with a focus on age and gender. We perform pre-training on facial images using the IMDb-WIKI dataset and subsequently classify both age and gender using unrestricted real-life facial data. To avoid over-fitting and enable the proposed model to generalise on the test images, we additionally chose a drop-out and data augmentation regularisation strategy. We demonstrate that optimised training hyper-parameters and well-planned network architecture produce better outcomes. The proposed technique achieved the 84.8 percent age group classification accuracy rate while the compared CNN2ELM model achieve the 52.3 percent accuracy rate. The suggested model demonstrates significantly better performance when contrast on the CNN2ELM model on the OIU-Adience dataset, especially in the context of age group classification and overall accuracy.