Potato Leaf Disease Classification using Convolutional Neural Networks
Yash Prashant Wasalwar, Kishan Singh Bagga, Vikash Kailash Joshi, Anuradha Joshi
Abstract
The vegetable Potato is quite familiar to all of us. After crops like rice and wheat, one of the most widely grown crops in India is the potato. But like other crops, Potato is also vulnerable to diseases. Two diseases namely Early blight and Late blight severely hinder the growth of potatoes. These diseases, if not discovered in its early stages, damage the crops, affects the production and apparently results in significant losses. The aim is to fasten the process of illness detection so as to increase the potato production by detecting the mentioned diseases at an early stage of Cultivation by using CNN, a Deep Learning based approach. The potato crop needs to be diagnosed by detecting and classifying leaves, whether they are healthy or infected and CNN algorithm is to be implemented for the same. By collecting the visual features of the leaves of distinct potato species, this study offers a deep learning-based approach for recognising the particular early and late blight infections in potatoes. Image processing was found the best way to detect and analyze the particular diseases as it helped us to extract the very essential information from the images of the leaves. Using a dataset obtained from subsidiary of plant village dataset, consisting of 3000 samples having 1000 samples each of healthy, early blight and late blight infected leaves, a custom model was trained with some advancements from our end to detect and categorize the leaves into healthy and diseased. The CNN based model gave an accuracy of 99% when prepared over 80% train data. The output shows that CNN outperforms all in the field of Potato disease classification.