Plant Disease Identification Using Transfer Learning

Abstract

Early detection and control of plant disease is of vital importance for better yield from crops. Plant disease can be identified from the leaves as the texture, color and spots are different from healthy leaves. Conventional method of observing the leaves require expertise. So development of plant disease detection using Deep Learning techniques such as transfer learning can help the farmers who lack expertise and resources to hire the expert. In this study, ResNet50 with Transfer Learning is used for disease identification of potato, tomato and corn. Performance of ResNet50 is compared with VGG16 and MCNN built and trained from scratch. ResNet50 achieved highest performance of 98.7% for plant disease identification. 16 classes of different plant diseases can be identified in the model. Work can be extended by training model on more classes.