A deep learning based crop disease classification using transfer learning

Abstract Healthy crops play crucial role in agriculture sector, which is backbone of our country. Unidentified crop diseases may lead to huge loss in agriculture sector and therefore rapid identification and detection of this is necessary. Correct identification and detection of the crop diseases may save the crops from being spoiled. Farmers cannot identify the disease just by observing the crop leaf, as the healthy crop and the effected crop appear same at the initial stages. This issue can be solved by implementing Deep learning models. The data collected for this study comprises of 20,639 images, which is taken from Plant village database. This research work is carried out by fine tuning the ResNet50 model with hyper parameters such as learning rate, mini-batch size and number of epochs. The proposed methodology achieves highest classification accuracy of 99.26 percent for fine tuning ResNet50.