Object recognition, particularly fruit classification, plays a crucial role in various fields, ranging from agricultural automation to digital marketplaces. This study proposes a fruit classification system based on RGB images, developed using a Convolutional Neural Network (CNN) architecture consisting of convolutional layers, pooling layers, fully connected layers, and dropout for model stability. The model was trained using the Adam optimization algorithm on an augmented dataset to enhance data variation and reduce overfitting. The resulting model achieved an average accuracy of 98%, demonstrating the reliability of CNNs in pattern recognition tasks. To enhance usability, the model was integrated into a graphical user interface (GUI) built with MATLAB R2023b App Designer, allowing users to add datasets, train the model, and predict new images without writing any code. The findings highlight that while the model performs well, its accuracy remains dependent on consistent image backgrounds; therefore, expanding the variety of fruit types and background conditions in the dataset is essential to improve the system's robustness in real-world applications.