|Keywords:||Convolutional Neural Networks, Transfer Learning, Detection, Recognition|
|Full text PDF:||http://e.buaa.edu.cn|
Agriculture is one of the major sources that interpose to economic growth in the country. However, diseases that attack the crops have given a little impact on agriculture production. Identification of plant leaf diseases is the preventive measure for the happened in the yield and the overall agriculture crop quantity. In the past times, the most widely adopted method by farmers was using the services of pathologists to find the defects from the plants. But as time changes the technology comes quickly in between to replace the old method, the older method was very costly, lengthy, and hectic. In recent years computer vision made drastic developments and now playing an important part in disease detection from plants like in other fields. There are extensive researches were conduct in this sector and different methods were adopted. Different techniques like machine learning (which include different algorithms like SVM, logistic regression, CART and KNN etc.) and deep learning (which include MLPNN, Backpropagation, RNN, and CNN, etc.). After reviewing the literature thoroughly, a most adopted and much precise algorithm CNN (ResNet34) is selected to complete the proposed model. Although there three major techniques to detect disease for the plant are root, stem, and leaves. Finding uncertainties using leave images is a very economic, fast, and easy way. Thus, three plants are selected to take their leaves pictures and to detect the various disease if presented any. The main idea is to find the defects first, then to choose the plant category and finally to classify that disease accordingly.