Disease Detection of Patato Using Deep Learning

Abstract

Computational analysis of plant disease is a challenging and interesting task in now a day. In this paper, I study on how to detect plant disease (potato late-blight) more accurately and give them solutions. Here I concerned about 3 different classes for classification. Several experiments are conducted on these collected dataset and extract feature of the classes. In this research we used deep learning based model (CNN – inception v3) for classification. First of all we train and fine tune our model. Then validate our model according to dataset. After 8000th iteration our classifier is able to classify different class accurately. In our experiment we measured up to 80% for all of the 3 classes. For first class late-blight disease, it gives accuracy level 87%, second class fresh potato detection is 86% and last if there have no potato in image then it also can detected, its accuracy is 87%.