Fresh and Rotten Fruit Classification

Abstract

Ensuring fruit quality through automation is becoming increasingly important in agriculture to maintain food safety and minimize post-harvest waste. This study proposes an automated system for identifying fresh and spoiled fruits by integrating Convolutional Neural Networks (CNN) and Random Forest Classifiers (RFC). The experiments were carried out using a publicly available Kaggle dataset containing 11,257 images of apples, bananas, and oranges. Image preprocessing steps such as resizing, normalization, and data augmentation were performed to enhance model robustness. The CNN model achieved a validation accuracy of 96.9% with an F1-score of 0.97, outperforming the RFC, which recorded 94.7% accuracy and an F1-score of 0.95. The CNN demonstrated superior feature extraction from spatial patterns, whereas RFC performance was constrained by its reliance on flattened image data. Furthermore, a web-based interface was implemented to allow users to upload fruit images and obtain classification results in real time. This research highlights the potential of combining traditional and deep learning techniques for intelligent fruit grading and real-time quality assessment in agricultural systems.