Transfer Learning based Tomato Ripeness Classification

The classification of the tomato has been very significant for organic farms that are generally concerned with executing the grading of tomato, based on its maturity feature. The conventional approach is time-consuming and costly as well. In an attempt to make optimal use of the rapid development of Neural Network's image classification and learning potential, the work focused on the selection and training of a pre-trained CNN is introduced to extract useful features from the collection of preprocessed data. The objective is to develop an automated tomato maturity estimator based on the color and shape that classify tomato images into eight distinct classes: Cherry Red-Round, Green-Round, Maroon-Round, Red-Flattened, Red-Heart, Red-Plum, Red-Semi Oval, Yellow-Semi Oval to eliminate the individual's interactions. Therefore, a revised version of the AlexNet CNN has been efficiently utilized to classify the tomato dataset. The proposed fine-tuned network achieved a validation accuracy of 100% using synthetic images.