Plant Disease Detection With Deep Learning And Feature Extraction Using Plant Village

This thesis addresses the issue of plant diseases detection using deep learning and feature extraction techniques.All the tests and experiments were carried out with an open sourced dataset Plantvillage.The mean idea behind our work it to compile three different deep learning models namely Resnet 50,Google Net,VGG16 and then find out which one of them fits the most regarding our classification problem.As we know,crop diseases are a crucial problem to the world's food supplies,in a world where the population count is around 7 billion people,with more than 90% not getting access to the use of tools or features that would identify and solve the problem.At present,we live in a world dominated by technology on a significant scale,major network coverage,high-end smartphones,as well as discoveries and improvements in AI.The combination of high-end smartphones and computer vision via Deep Learning has made possible what can be defined as ?smartphone-assisted disease diagnosis?.In the area of Deep Learning,multiple architecture models have been trained,some achieving performance reaching more than 99.53%.Previous research was done separately on every model,each of them generating their own results.However,in our research we put together three previously tested deep learning models(Resnet 50,Google Net,VGG16)and two classifiers(SVM and KNN)using the state-of-the art technique algorithms in order to compare the results obtained and find out which of the models is more accurate and able to solve better the plant disease classification problem.In this study,we answered that question as our results proved that Resnet 50 is the best model for feature extraction;VGG16 is preferred when doing transfer learning;SVM has proven itself to be the best classifier compared to KNN.