Apple and Grape Leaf Diseases Classification using Transfer Learning via Fine-tuned Classifier

Abstract

Listen

Translate article

Plant leaf endemics are part of human life. Any changes in crops yield leads to shortage of food. Apple and Grape fruits are most profitable and also prone to many diseases. Apple trees often suffers from Scab, Black rot, and Cedar rust diseases and Grape plants suffers from Black-Measles, Black-rot, and Leaf-Blight diseases. The productivity of Apple and Grape depends on early detection and diagnosis of diseases. The various parts of plants such as leaf, trunk, seed, blossom and fruit growth gets affected. Identification and classification of these endemics requires presence of farmer or plant pathologists. Manual diagnosis of disease might lead to misidentification and inappropriate use of pesticides and also consumes lot of time. Plants suffer from incorrect pesticide usage to diagnose endemics. There is a need for artificial ways in classifying diseases. In this work, a fine-tuned VGG-16_Network is proposed to classify 8 different categories of apple and grape leaf together. This model is capable of categorizing separate diseases of Apple and Grape leaves which reduces the training time and identifies the diseases accurately. Here fine tuning of VGG-16_Network is done by removing last layer of existing network and appending a new output layer to it. The model's training time is reduced by adopting transfer learning, that is, only the last layer is trained and the remaining VGG-16_Network layers are not trained with apple and grape leaves dataset. Our experimental results shows that, the training parameters are drastically reduced by 98.9% compare to training entire model. Therefore, the proposed Deep-convolutional neural network provides exceptional result in classifying apple and grape leaf diseases with an accuracy of 97.87%.