A Machine Learning Approach to Crop Disease Detection using Image Processing and Feature Extraction

One of the major issues faced in agricultural fields worldwide is detection of plant illness. Identification of such diseases in the early stages is crucial to improving overall productivity and reducing production loss. Farms that contain a variety of crops cannot be tracked manually for diseases as it requires skilled labor and it is resource intensive. The thesis introduces an accurate and effective method for detection of plant illness which uses Convolutional Neural Network (CNN) architecture (VGG style). The proposed system can detect 26 different diseases of 11 plants with 98% accuracy.