Morphopod: A Convolutional Neural Network Architecture for Predicting Human Age Based on Footprint Analysis

Abstract

Identifying human age based on Footprints is a complex task, primarily due to the absence of comprehensive datasets. To tackle this issue, we have developed a dataset divided into four distinct categories. Our initial strategy involves using pre-trained models to apply transfer learning, training on RGB images of human footprints from the dataset for classification. This paper introduces Morphopod, a custom Deep CNN architecture optimized to outperform tradi- tional pre-trained models. By addressing challenges like data imbalance, which often under- mines the success of pre-trained systems, Morphopod offers a more reliable solution. It employs Gaussian filtering applied to annotated images as inputs, significantly reducing environmental interference, noises, and achieving better accuracy. This architecture serves as a foundational step for solving footprint-based human identification tasks and holds potential for various other applications in this domain.