Quality classification of tomato plant in field conditions using EfficientNet deep learning model

The appropriateness of the agricultural economy is very effective in sustainable food security. The appearance and shape of agricultural products change in different periods. The correct classification of the product in terms of quality after harvest affects the economy of farmers. Today, deep learning classifiers have greatly contributed to the correct classification of product quality. But the database challenges and the same conditions of the database in the training and testing phase affect the classification accuracy. The purpose of this article is to classify the quality of tomatoes in the challenging conditions of the database, including crowded backgrounds, noise in the image, leaves of the same color as the fruit in the image, and the similarity of growth stages. For this purpose, 3 databases with different challenges have been used in the stage of classification training and testing. In this article, the aim is to classify the quality of tomatoes into 3 classes ripe, unripe ,and semi-ripe using Efficientnet deep learning classifier. According to the conditions of the database, the first three processes of noise removal, image contrast improvement ,and image segmentation have been applied to the images. The results of the evaluation of the proposed method show the proper performance of EfficientnetB5.