The categorization of papaya fruit varieties is important within the agricultural industry. There is a need for a system that can accurately evaluate the different varieties of papaya fruits to save labor expenses associated with discarding specific types of papaya fruits throughout the packaging and distribution process. The present work introduces a novel approach for the categorization of papaya fruit varieties with the aim of enhancing efficiency and accuracy. The proposed method comprises a series of interconnected steps. Once the data on papaya fruits are collected, they undergo preprocessing techniques, such as color uniformization, image scaling, augmentation, and image labelling. Subsequently, the AlexNet architecture was employed, incorporating 08 layers, comprising 05 conv layers and 03 fully connected layers. Simultaneously, the process of transfer learning and fine-tuning of the convolutional neural network was executed. During the concluding phase, a SoftMax classifier was employed to perform the ultimate classification. Detailed simulations were conducted using self-generated datasets. The proposed model demonstrated the highest classification accuracy of 98.0%. Furthermore, our developed model exhibits computational efficiency with an average processing time of 8 ms to generate the ultimate classification outcome. The obtained results enhance the comprehensibility and reliability of convolutional neural network (CNN) models, thereby offering valuable insights for the prospective utilization of deep learning techniques in agricultural fields.