Machine learning based model for leaf diseases detection

Healthy plant leaves are essential for increasing the yield whereas diseased leaves cause variable loss in the overall production. Identification and classification of diseases in plant leaves in the early stage is a comprehensive initiative designed to automate the prediction and classification of disease affecting plant leaves through the utilization of machine learning techniques. The primary objective is to revolutionize disease detection within agriculture by providing an accurate, efficient, and accessible solution for both farmers and experts. This paper offering a comprehensive solution for leaf diseases prediction using convolution neural networks and transfer learning techniques implemented in python with Tensor flow, Keras libraries and flask framework. The workflow encompasses data augmentation, model creation, training, testing and validation. The dataset, consisting of images collected from public dataset subjected to data augmentation to improve model performance. The pre-trained image dataset is employed as the backbone for feature extraction, trailed by a custom dense layer for prediction of the tumor. The model is trained and evaluated over 15 epochs achieving a remarkable accuracy of approximately 95.84% in training and in testing or validation the accuracy ranges around 87.0%. The model is expert to accept single leaf image from the image database for the estimate of the diseased leaf or healthy leaf.