Performance Evaluation of Retrained CNN Models for Grape Leaf Disease Identification

In smart agricultural systems, the identification of diseases from leaf images plays a crucial role in achieving prompt diagnosis and enhancing crop production. Convolutional Neural Network (CNN) models based on deep learning have demonstrated substantial improvements in the accuracy of disease detection. This study focuses on the detection of grape leaf diseases using a retraining approach applied to standard CNN models. The evaluation encompasses popular CNN architectures such as VGG, ResNet, Xception, Inception, DenseNet, and MobileNet for classifying grape diseases. The research involves experimentation on two established datasets and the compilation of a new dataset, named Dataset 3, to expose the model to diverse training scenarios. The findings underscore the pivotal role of data quality in model performance. Significantly, the models exhibit excellence on Dataset 3, showcasing potential for early detection of grape diseases. This research contributes to the advancement of agricultural technology, influencing precision agriculture and promoting sustainability.