Image Based Age Prediction and Gender Classification Using Convolutional Neural Network: A Comprehensive Analysis

Gender recognition from facial photographs is a primitive task in various fields including security, marketing, and in many other fields. This aims to create a suitable gender recognition model on a Kaggle dataset having more than 20,000 images. Training on large dataset gives us a more accurate results. Transfer learning methods were used to accelerate training and increase generalization using ResNet50, a deep learning model developed on Convolutional Neural Networks (CNN).In this paper we show that by learning representations through the use of deep-CNN, a significant increase in performance can be obtained on these tasks. The model achieves both, accurate classification and minimized overfitting, through regularization.. Results demonstrate that the proposed model achieves accuracy in both gender and age detection task, even under challenging conditions such as background with light ,pose, person with objects. Performance evaluation for gender classification was achieved by estimating the precision which was 92.91%, recall which was 96.78% and F1-score which was 94.23%. For age prediction Means Absolute Error (MAE) was estimated under various age groups and overall average MAE obtained was 30.