Classification of Kinalabasa Tomato Using Convolutional Neural Network

Determining the freshness of tomatoes is essential for evaluating their quality, impacting both consumer satisfaction and the economic benefits for farmers. Freshness is typically assessed by outer appearance, including color, size, and shape, with skin color indicating ripeness and influencing selling price. Properly assessing freshness also helps determine the shelf life of stored tomatoes. This study introduced an affordable and straightforward technique for classifying agricultural commodities using image processing, focusing on Kinalabasa tomatoes and classifying them based on color features into three categories. The study employed the MobileNetv2 architecture for training and testing the dataset, aiming to improve the classification of Kinalabasa tomatoes through deep learning techniques. The study evaluated the model's performance using metrics like accuracy, precision, sensitivity, and specificity. The optimal configuration for classifying Kinalabasa tomatoes was determined to be 10 epochs and a learning rate of 0.001. MobileNetv2 achieved a high accuracy rate of 94%, demonstrating its effectiveness in classifying tomatoes, though the specificity rate of 64% indicated some room for improvement in identifying negative instances. Comparing MobileNetv2 with ShuffleNet architectures will provide further insights into their respective effectiveness and performance in this classification task.