Desarrollo de un sistema de visión artificial para la detección de la plaga Botrytis cinerea en el cultivo de uva

The present study aimed to develop a computer vision system for the detection of the pest Botrytis cinerea in grape crops (Vitis vinifera). Applied research with an experimental design was carried out, beginning with the capture of 750 images in a vineyard in Cascas, La Libertad, Peru. Of these, 600 were classified and labeled in Roboflow, differentiating between healthy and infected grapes. This dataset was subsequently used to train the YOLOv11 model in Visual Studio Code using Python. The model reached an augmented database of 1,560 images and was trained for 100 epochs. The model's results showed a sustained improvement in the loss curves, showing a constant reduction and an overall accuracy of 92.5% under uncontrolled conditions. In addition, a mobile camera was integrated for real-time detection, validating its applicability in the field. Although limitations were identified when faced with complex backgrounds or subtle symptoms, the system proved to be effective, accessible, and replicable. Concluding that this artificial intelligence-based solution represents a viable technological alternative for phytosanitary monitoring, with the potential to be scaled to other high-value crops.