Explainable Age and Gender Classification Using CNN with SHAP Interpretability on Human Facial Images

Abstract

Many practical applications, for example, social media analysis, targeted marketing, human–computer interactions, and security systems, become possible through the reliable and accurate identification of a person, in terms of their gender and age, based on a photo of their face. The paper involves giving an end-to-end solution for online deployment that utilizes convolutional neural networks (CNNs) and explainable artificial intelligence (XAI) methods to classify age and gender in real time. The model uses Adience data and data augmentation to learn and cope with the imbalance in classes of age and gender. The CNN design uses convolutional, pooling, and fully connected layers to identify features, compress the input, and connect to the eight-age and gender-category classification target output branches. Several metrics may be used to gauge the efficacy of a model, including recall, F1-score, accuracy, and precision. Visualizing the impact of face traits on prediction results is possible using the SHapley Additive exPlanations (SHAP) approach, which does not sacrifice model quality or user trust. The web platform is built on a Flask framework to enable users to input facial images to the system and get an immediate prediction with explanations of the prediction provided in SHAP format. The usage of XAI proves the effectiveness and explainability of artificial intelligence use in the real-life context. Through experimental results, it is found that the model is highly accurate and explainable and hence is suitable to be used in real-life applications.

Received: 31 October 2025 | Revised: 14 July 2026 | Accepted: 5 August 2026

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

The data that support the findings of this study are openly available on Kaggle at https://www.kaggle.com/datasets/ttungl/adience-benchmark-gender-and-age-classification.

Author Contribution Statement

Patel Alpaben Rameshbhai: Conceptualization, Methodology, Software, Validation, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Project administration. Amisha Shingala: Formal analysis, Resources, Data curation, Writing – original draft, Visualization, Supervision.