Plant Disease Classification using Deep Learning Approach (VGG19)

Plant disease classification is a prime research area in precision agriculture. Artificial Intelligence (AI) based automatic disease prediction using machine learning or deep learning models is gaining momentum in recent years to protect sensitive economy crops such as tomato and potato blight at the early stages of the fungal infection. One of the challenges in implementing deep learning framework is limited dataset availability. In this proposed work, VGG-19 a Convolution Neural Network (CNN) model are being used to predict the disease classes based on transfer learning approach. Early detection of diseases aids in minimizing the yield losses by 30 % and CNN based models such as VGG-19 plays a significant role in extracting the low-level spatial information. This VGG-19 is pre-trained on ImageNet dataset with more than thousand classes and CIFAR 100 dataset. This VGG-19 is like VGG-16 with three additional fully connected layers. The knowledge learnt from the benchmarked dataset is applied to the ten classes of diseased tomato leaves. The accuracy metrics are plotted, and the highest accuracy obtained is 94 %. The confusion matrix results are also obtained to validate the performance.