Abstarct: This thesis investigates and classifies fruit images using deep learning techniques. Fruits, as one of the primary nutritional elements for humans, play a crucial role in meeting nutritional needs and promoting human health. This research evaluates the accuracy and efficiency of fruit image classification using pre-trained models, including ResNet50 and VGG16, along with attention modules. The results demonstrate that applying these techniques improves classification models' performance and efficiency. The developed model in this study has significantly increased the accuracy of fruit image classification to a range between 89% and 100%, which is a notable improvement compared to previous methods that had an accuracy of around 85%. Fruits such as apples, bananas, and oranges, due to their richness in vitamins, nutrients, and antioxidants, are always considered an essential part of the human diet. Due to their variety in taste and nutritional capabilities, whether fresh, dried, or processed into products like juice, compote, and jam, they are always available and widely consumed. In this research, the data were first collected and preprocessed. Then, deep learning models were designed and implemented using the ResNet50 network and Attention modules. Finally, the performance of these models was evaluated and analyzed. The results indicate a significant improvement in speed and accuracy in the detection and classification of fruit images using the ResNet50 network and Attention modules.