Weapon Classification with Xception: An Efficient Deep Learning Approach

Intelligent security systems require robust capabilities for weapon detection and classification. This study employs the Xception architecture, leveraging its unique features such as depth-wise separable convolutions and hierarchical feature extraction. A diverse dataset encompassing weapon and non-weapon entities is meticulously annotated and preprocessed. The Xception model is fine-tuned to effectively identify weapons, achieving notable results with an accuracy of 98.3%, precision of 98.50%, recall of 99.49%, and f1-score of 98.99%. Comparative evaluation with YOLOv8, ResNet50, and ResNet18 confirms Xception's superiority in recognizing subtle weapon patterns. Rigorous evaluation on relevant datasets provides empirical evidence of the model's effectiveness in real-world scenarios. These findings underscore Xception's practical viability in enhancing security applications, contributing substantively to the discourse on advanced neural network architectures' role in addressing contemporary security challenges. This research emphasizes the importance of innovative approaches in augmenting security systems' capabilities to ensure public safety.