DETECÇÃO AUTOMÁTICA DE CERCOSPORA EM FOLHAS DE FEIJÃO CAUPI: ABORDAGEM BASEADA EM REDES NEURAIS CONVOLUCIONAIS

This study evaluated the performance of three convolutional neural network models – LeNet, ResNet50, and EfficientNetB0 – in the task of classifying cowpea leaves as healthy or infected by cercospora. The data were collected from cowpea-producing regions in Piauí and used to train and test the models. The results showed that LeNet achieved an excellent balance between performance and efficiency, with 91.67% accuracy and lower processing time. ResNet50 achieved the highest accuracy (95.83%) but at a higher computational cost, while EfficientNetB0 showed similar performance to LeNet but required more resources. In conclusion, the study suggests that LeNet is the most suitable model for practical applications in this context.