Mobile application for the detection of diseased apples: Comparison between SOM and CNN networks

Sahuma Jurado G.D.; Inca Huamaní B.O.; Huarote Zegarra R.E. · Proceedings of the LACCEI international Multi-conference for Engineering, Education and Technology (2025) · DOI: 10.18687/leird2025.1.1.264

Resumen

Detecting diseases in apple trees is a critical challenge for modern agriculture, as it leads to production losses and excessive use of agrochemicals. This study developed a mobile application to identify whether an apple is healthy or diseased, from a total sample of 5,000 photographs of apples grown on the outskirts of Lima, Peru. The images were captured in real field conditions, obtaining photographs of fresh, rotten, dull, and differently colored apples, etc. The SOM neural network achieved 94% accuracy The SOM neural network was shown to achieve 94% accuracy compared to 92% for the CNN, with processing times of less than 10 seconds, maintaining greater robustness in the face of variations in lighting and the environment. Although CNN showed advantages in computational efficiency, its lower accuracy determined the final selection of the SOM architecture for the application. The proposed solution operates on standard mobile devices and offers small producers an accessible and reliable tool for post-harvest quality control, contributing to the reduction of economic losses and the adoption of low-cost agricultural technologies. © LEIRD 2025.All rights reserved.

Instituciones: Universidad Nacional Tecnológica de Lima Sur