Identificación de tipos de residuos sólidos para optimizar el proceso de reciclaje mediante técnicas de aprendizaje profundo

Abstract

Abstract: Improper solid waste management represents a significant environmental challenge, causing pollution and waste of resources that directly affect the planet's sustainability. To address this problem, this study focused on the automatic identification and classification of solid waste using deep learning techniques. To do this, the Kaggle TrashNet database was used. Using these images with different types of waste, three neural network architectures were tested. These were obtained through analysis of the state-of-the-art: AlexNet, YOLOv8, and YOLOv11. Different tests were conducted with evaluation metrics to measure their performance. In these, YOLO version 11 demonstrated the best performance of the three, with a MAp of 96.8%, precision of 90.9%, and recall of 92.0%. The RoboFlow platform was used for image labeling and implementation of the trained model. These results highlight the effectiveness of YOLO-based models for waste classification, contributing to the development of technological solutions that optimize recycling and promote more sustainable materials management.