Development of an Artificial Vision Algorithm for the Classification of Tomatoes by Quality in the Markets

León León R.A.; Astete Zamalloa I.; Villalta Suarez K.N. · Memorias de la Conferencia Iberoamericana de Complejidad, Informatica y Cibernetica, CICIC (2025) · DOI: 10.54808/cicic2025.01.308

Resumen

The objective of this project was to develop a computer vision algorithm for the automatic classification of tomatoes by quality in markets, with a minimum precision of 95%, using the YOLOv8 architecture. To achieve this, a dataset of 2,452 images of tomatoes under varying lighting conditions, angles, and quality levels was selected and preprocessed via the Roboflow platform. The model training phase was conducted in Google Colab, utilizing an A100 GPU to optimize processing times. YOLOv8 was configured in a Python environment using PyTorch to carry out the network training. Key hyperparameters, such as batch size, learning rate, and the number of epochs (set to 350), were adjusted. During training, metrics such as loss and validation accuracy were monitored, with adjustments made to optimize performance. The model achieved a 98% accuracy in tomato classification, confirming the effectiveness of YOLOv8 for automatic quality classification applications for tomatoes. Copyright 2025. © por el International Institute of Informatics and Systemics.

Instituciones: Universidad Privada del Norte