Implementation of YOLO Neural Networks for the Detection of Box Opening in Packaging Systems

León León R.A.; Aguilar Zavaleta J.I.; Chávez Vásquez D.M.; Briones Flores E.L.; Mejia Rubio G.H.; Huaringa Roldan J.A. · Memorias de la Conferencia Iberoamericana de Complejidad, Informatica y Cibernetica, CICIC (2025) · DOI: 10.54808/cicic2025.01.276

Resumen

The aim of this study is the implementation of YOLOv8 neural networks to enhance the automatic detection of unauthorized openings in boxes within parcel delivery systems. The methodology included the creation of a varied dataset from proprietary images and a public database, managed and labeled in Roboflow, ensuring adequate representation of real-world lighting conditions and viewing angles. YOLOv8, an advanced version of the algorithm, was used in Google Colab for training, utilizing image preprocessing and postprocessing techniques that improved the model’s accuracy and reliability in detecting openings. The results showed a 98% accuracy in detecting openings. The application of YOLOv8 for detecting unauthorized openings in boxes provides a modern and precise solution to enhance security and quality control in the supply chain. Copyright 2025. © por el International Institute of Informatics and Systemics.

Instituciones: Universidad Privada del Norte