Identification of Fruit Disease Using Instance Segmentation

Agricultural farms that have undergone commercialization are continually looking for methods to minimize manpower without sacrificing output. This work proposes and experimentally validates a method for detecting and classifying apple fruit disease. This system takes a picture of an apple fruit as input and determines if it is infected or not. Farmers would be able to spot illnesses in fruits using this approach. Instance segmentation an advancement of image segmentation is used to identify every pixel belonging to an instance of an object and it detects distinct object. Normal apple fruit diseases are considered and then marked in the existing system. There is no algorithm used to find the type of diseases affected. RNN was used for the detection and segmentation of disease in apple fruit. In this work CNN algorithm is proposed which is used for identification of disease in the apple fruit. Few types of fruit diseases, namely bitter rot, sooty blotch and powdery mildew images are used for this approach. Instance segmentation is performed using CNN algorithm to identify affected parts in the apple fruit. Compared to the RNN algortihm, CNN algorithm shows better accuracy for the identification of disease in the apple fruit by applying instance segmentation.