Deep Learning-Based Tomato Feature Detection and Quality Assessment in Horticulture

This study focuses on the application of the CNN for real-time detection and sorting of tomato fruits with reference to their maturity and quality. A modified ResNet50V2 architecture was employed, achieving a classification accuracy of 95.36% across four categories: Unripe tomatoes, tomatoes at the right maturity, over ripe tomatoes, and tomatoes with blemish. There was a total of 3345 high resourced images which were preprocessed through a process that included augmentation and normalization so as to prevent the models from being overfit. To fine tune model and speed up the convergence, different types of optimizers such as Adam and Adamax has been used. The purpose of this work is to evaluate the possibility of Tomatoes’ quality categorization using CNNs. This work has pointed out that deep learning technologies especially CNN could advance agriculture practices by boosting crop quality and minimizing post-harvest loss. This research emphasizes the application of deep learning techniques in the existing agri-food systems to enhance the efficacy of crop management and the quality of yields.