Development of a Machine Vision Algorithm for the Recognition of Firearms and Knives

León León R.A.; Linares Valdiviezo X.B. · Memorias de la Conferencia Iberoamericana de Complejidad, Informatica y Cibernetica, CICIC (2025) · DOI: 10.54808/cicic2025.01.283

Resumen

This project developed a computer vision algorithm based on YOLOv8 for the recognition of firearms and bladed weapons, aiming to improve public safety through fast and accurate detection. The methodology began with the collection of 4,620 images of weapons, manually labeled. The model was then trained in Google Colab using an A100 GPU, optimizing hyperparameters such as batch size and learning rate over 250 training epochs. To facilitate real-time detection, a Python script was implemented in Visual Studio, allowing the loading and analysis of new images. The results show a high precision of 98% in weapon identification, confirming the algorithm's effectiveness in security environments. This YOLOv8-based system provides an advanced solution for violence prevention in public and private spaces, enabling a rapid response and strengthening security. Copyright 2025. © por el International Institute of Informatics and Systemics.

Instituciones: Universidad Privada del Norte