Abstract

MARTINEZ, Fatima; ROMAINE, James Brian; CARDONA, Adrián  and  MILLAN, Pablo. Computer Vision-Based Detection of Lettuce Health. Rev. ciente. UCSA [online]. 2025, vol.12, n.1, pp.3-13. ISSN 2409-8752.  https://doi.org/10.18004/ucsa/2409-8752/2025.012.01.003.

Early identification of plant health issues is essential for efficient and sustainable agricultural management. However, traditional visual inspection methods are not feasible on a large scale. Therefore, this study proposes an automated system for detecting the condition of lettuce crops using the YOLOv5 object detector. The model was trained on a dataset comprising 144 images of lettuces, captured at the BIOAlverde farm (Seville, Spain), and classified into two categories: healthy and unhealthy. The images were manually annotated, and the model was trained using data augmentation techniques, hyperparameter tuning, and cross-validation to improve accuracy. The results obtained demonstrate the high performance of the system. For healthy lettuces, the model achieved a precision of 97.9%, a recall of 99.3%, and a mean Average Precision (mAP) of 99.5%; while for unhealthy lettuces, precision reached 95.8%, recall 100%, and mAP 99.5%. In conclusion, the proposed system constitutes an effective tool for the automatic detection of lettuce condition, contributing to the advancement of precision agriculture through the optimisation of resource use and the timely identification of crop issues.

Keywords : YOLO; Lettuce detection; Computer visión.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )