Single-Level Fusion for Enhancing Meat Quality Classification with Explainable AI

Growing the global meat market, especially in the Asia-Pacific region, also brought a desire to continue providing quality meats, and this implies the need to invest in a high-quality meat supply to avoid health issues associated with spoiled meat like stomach upsets and food poisoning. This study proposes a state-of-the-art deep-learning model of a single fusion level of ConvNeXtTiny and DenseNet169 networks for categorizing meat as fresh, half-fresh, and spoiled. Pivotal to this study is the use of Grad-CAM++ for explainable AI (XAI) that offers finer localization by generating high-resolution heatmaps that focus on the important regions in images. This approach enhances the interpretability of the model’s decision-making process in a way that enables users to determine which of the features contributes to classification results easily. The model, trained on a comprehensive dataset of 10,931 images, achieved a notable testing accuracy of 97.46%, outperforming recent studies. Therefore, besides demonstrating the effectiveness of the chosen XAI techniques, this study contributes to enhancing the safety and quality of meat products by providing insightful explanations of the operations of the AI system.