A Novel Classification Model of Date Fruit Dataset Using Deep Transfer Learning

Date fruits are the most common fruit in Middle East and North Africa. There a wide variety of dates with different types, colors, shapes, tastes, nutritional values. Classifying, identifying, recognizing would play crucial role agriculture, commercial, food, health sectors. Nevertheless, there is no or limited work to collect reliable dataset for many classes. In this paper, we collected date by picturing from primary environments: farms shops (e.g., online local markets). The combined unique due multiplicity items. To our knowledge, contains same number classes natural environments. has 27 3228 images. experimental results presented based on five stages. first stage applied traditional machine learning algorithms measuring accuracy features pixel intensity color distribution. second deep transfer (TL) model select best classification. third stage, feature extraction part was fine-tuned applying retrained points retraining point. fourth fully connected layer achieve classification configurations model. fifth regularization best-selected where validation reached 97.21% test 95.21%.