Plant disease is still a big challenges in agricultural production across the world. Manual examination alone can be time-consuming, labour-intensive, and error-prone, and is hard to be extended to large- scale cultivation. To overcome these constraints, we propose a CNN-based method for automated early detection of plant diseases. The model is trained on a large-scale dataset of more than 87000 images of 38 classes of healthy and infected plant leaves. Prior to training, a structured preprocessing pipeline including resize, normalize, and data augmentation on images to increase variability within samples and reduce the risk of overfitting is applied to all images. The CNN model consists of convolution and pooling layers with a final fully connected layer and the optimization is performed with Adam optimizer. Experimental results demonstrate that the proposed method yields an accuracy of 96%. As consequence of the comprehensive augmentation strategy, the model is also extended to unsee samples and obtains better performance than other famous architectures like AlexNet and VGG16. These results underscore the ability of deep learning to facilitate precision agriculture through early disease detection, which allows farmers to make better management decisions and improve yield and farm sustainability.
Vanguru et al. (Fri,) studied this question.