A pilot study at the University of Catania investigated the use of deep learning through the free Lobe AI software for the classification of gunshot wounds (GSW) in forensic pathology. They analyzed four categories: GSW in general, entry/exit wounds, wounds by pathological firing range, and wounds by ammunition type. Training images came from the GSW forensic atlas, test images from Catania cases, and healthy skin photos as controls. The process included four phases: training, validation, testing, and data analysis with parameters precision, accuracy, recall, F1 score, and specificity. The results showed encouraging data, where many parameters exceeded the human limit and achieved the highest predictive values in some areas, including a comparison with previous studies with 98% accuracy. Limitations included a lack of training data and the need for specific software. Strengths were the four categories and the use of an "intact skin" control. The study recommends extension to multicenter research with a larger sample.