Automated solid waste classification through computer vision and the YOLOv8 model for environmental management

Authors

Thomás Ricardo Borja Saltos
Laboratorio de Investigación en Inteligencia y Visión Artificial, Departamento de Informática y Ciencias de la Computación, Escuela Politécnica Nacional, Quito 170525, Ecuador. ROR: https://ror.org/01gb99w41
https://orcid.org/0000-0002-5475-1016
Claudio Xavier Borja Saltos
Departamento de Desarrollo de Software. SOFTECAPPS S.A.S. Guaranda 020101, Ecuador.
https://orcid.org/0009-0008-6938-9399
Jorge Wilson Tamami Pachala
Facultad de Ciencias de la Educación, Universidad Estatal de Bolívar. Guaranda 020150, Ecuador. ROR: https://ror.org/005cgg117
https://orcid.org/0000-0002-3470-7894
Cristhian Geovanny Villamarin Arroba
Unidad de Posgrados. Universidad Estatal de Milagro. Milagro 091050, Ecuador. ROR: https://ror.org/00gd7ns03
https://orcid.org/0009-0000-1905-4015
DOI: https://doi.org/10.63804/mtc.v2i1.e1

Synopsis

Sustainable urban solid waste management faces significant challenges due to the low efficiency of manual classification processes and the limited recovery rate of recyclable materials. This study aimed to evaluate the effectiveness of a computer vision model based on the You Only Look Once (YOLOv8) architecture for the automated identification and classification of waste. The methodology involved processing an optimized version of the TrashNet dataset, consisting of 2,390 images after a rigorous cleaning and balancing procedure. The dataset was categorized into five fundamental classes: cardboard, paper, plastic, metal, and glass. Transfer learning techniques and stratified segmentation were employed to ensure robust training within a high-performance computing environment. Experimental results demonstrated superior performance, achieving an overall validation accuracy of 0.94. Precision, sensitivity, and F1-score metrics revealed high inter-class consistency, maintaining operational stability under variations in lighting and visual noise. In the discussion and impact phase, technological applicability was validated through the deployment of the trained model in functional desktop and web-based applications. These tools enabled real-time processing without reliance on external servers, facilitating the decentralization of environmental monitoring. It is concluded that the integration of single-stage convolutional neural networks represents a scalable and computationally efficient solution, with significant potential to transform circular economy systems and optimize resource recovery in urban and industrial contexts.

Author Biographies

Thomás Ricardo Borja Saltos, Laboratorio de Investigación en Inteligencia y Visión Artificial, Departamento de Informática y Ciencias de la Computación, Escuela Politécnica Nacional, Quito 170525, Ecuador. ROR: https://ror.org/01gb99w41

Holds a B.Sc. in Electronics and Telecommunications Engineering and an M.Sc. in Software Engineering with a specialization in Security from Escuela Politécnica Nacional. He is a full-time lecturer in the Software Development program at the Faculty of Systems Engineering of the same institution. His academic work focuses on software engineering, cybersecurity, and the experimental evaluation of artificial intelligence-based systems, emphasizing reproducible methods and applied technological solutions.

Claudio Xavier Borja Saltos , Departamento de Desarrollo de Software. SOFTECAPPS S.A.S. Guaranda 020101, Ecuador.

Holds a degree as Information Technology Technologist with a specialization in Systems Analysis and a degree as Software Development Technologist from Instituto Tecnológico El Libertador (Ecuador). He works as a software developer and serves as Sales Manager at SOFTECAPPS S.A.S. His professional experience focuses on application implementation, software integration, and the deployment of end-user-oriented systems, combining technical execution with management-oriented decision-making.

Jorge Wilson Tamami Pachala, Facultad de Ciencias de la Educación, Universidad Estatal de Bolívar. Guaranda 020150, Ecuador. ROR: https://ror.org/005cgg117

Holds a B.Ed. in Education with a specialization in Educational Computing, a postgraduate specialization in Data Communication Networks, and an M.Sc. in Business Informatics. He has taught in multiple educational institutions and in higher education, and has held public administration roles in Bolívar Province, including local and provincial leadership positions and educational district management. His background integrates education, institutional management, and applied information technologies.

Cristhian Geovanny Villamarin Arroba, Unidad de Posgrados. Universidad Estatal de Milagro. Milagro 091050, Ecuador. ROR: https://ror.org/00gd7ns03

Holds a B.Ed. in Educational Computing from Universidad Estatal de Bolívar, a degree as Software Development Technologist from Instituto Tecnológico El Libertador, and an M.Sc. in Information Technologies from Universidad Estatal de Bolívar. He works as a lecturer and researcher with interests in applied computing, data science, and software development. His work emphasizes computational approaches to support educational and technological improvement initiatives.

Cover MTC V2i1

Downloads

MTC V2i1 I 2026

Volume

Challenges of contemporary society

Pages

18

Published

February 28, 2026

Series

Market, Technology, and Citizenship

Online ISSN

3103-117X

Categories

Copyright (c) 2026 Thomás Ricardo Borja Saltos, Claudio Xavier Borja Saltos , Jorge Wilson Tamami Pachala, Cristhian Geovanny Villamarin Arroba, Mercedes Anabel Monar Verdezoto, Hernán Arturo Rojas Sánchez, Simeón Ángel Zavala Verdezoto, Alisson Daniela Ocampo Paño, Mario Manuel Paredes Lopez, Erika Anabel Duche Manobanda, Jesús Ernesto González Torres, Alexander Ufredo Alegria Chavez, Diana Magali Alegría Camino, Verónica Elizabeth Sánchez Aguiar (Autor de capítulo)

License

Creative Commons License

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

How to Cite

Automated solid waste classification through computer vision and the YOLOv8 model for environmental management. (2026). In Challenges of contemporary society: Vol. V2i1 (p. 18). Editorial Unión Científica. https://doi.org/10.63804/mtc.v2i1.e1