Optimizing Deep Learning Models for Aflatoxin Detection in Agricultural Products: A Case Study of Groundnuts.

Abstract

This study presents an automated deep learning-based classification model for aflatoxin detection in groundnuts, addressing the limitations of conventional manual inspection methods, which are often time-intensive and error-prone. Leveraging the Inception-ResNet-V2 deep learning architecture, the model classifies groundnuts into four distinct categories: healthy, moldy, pest-infested, and those exhibiting physiological disorders. A comprehensive dataset comprising 226 healthy, 236 moldy, 191 pest-infested, and 160 physiological disorder samples was utilized for training, validation, and testing. Model performance was evaluated using multiple metrics, including accuracy, precision, recall, F1-score, and the area under the receiver operating characteristic (ROC) curve (AUC). The proposed model achieved an overall accuracy of 99.29%, with precision and recall values of 100% and 98.44%, respectively. Notably, the moldy category exhibited an AUC of 1.00, underscoring the model’s exceptional capability in distinguishing visual patterns and automating classification tasks. Despite these good results, the study highlights the need for future research to incorporate a broader range of agricultural products to enhance model generalizability. The deep learning model developed improves aflatoxin detection, reducing reliance on subjective manual inspections and enhancing food safety practices. This research offers a novel AI-driven solutions in agricultural quality assessment and food safety management.