CHALLENGE FOR THE CLASSIFICATION OF LABELED RECYCLABLE MATERIALS FOCUSED ON GLASS AND PLASTIC USING ARTIFICIAL INTELLIGENCE AND COMPUTER VISION(Desafio para a classificação de materiais recicláveis rotulados com foco em vidro e plástico utilizando inteligência artificial e visão computacional)

Resumo

https://doi.org/10.5281/zenodo.20214267

This work presents the development of an automated system for identifying and classifying labeled recyclable materials, with emphasis on plastic and glass, using computer vision and artificial intelligence techniques. Considering the growing volume of urban solid waste and the low efficiency of manual sorting methods, the proposed solution aims to improve the accuracy, speed, and safety of the separation process.

For this purpose, a dataset in YOLO format was prepared and processed through the Roboflow platform, and the YOLOv11 model was trained in two distinct versions, enabling comparative performance analysis. The results show significant improvements, especially in the “glass” class, whose recall increased from 66.67% to 89.66%. The comparison between versions highlights advancements in stability and overall model performance. The developed system demonstrates potential to optimize recycling processes, reduce operational costs, enhance the quality of recovered materials, and foster sustainable practices aligned with the circular economy.