POTATO DISEASE CLASSIFICATION USING GRADIENT BOOSTING

Abstract

Potatoes are one of the widely consumed foods throughout the world. Usage of potatoes increases day by day. India is the second largest country in producing potatoes. It is good if we predict the disease earlier. By this wastage of potatoes decreases. Most of the potato disease can be predicted based on condition of leaf. Potato disease are of 2 types – Early blight and Late blight. Dataset is taken from Kaggle website which contains 2000 pictures of healthy and unhealthy potato’s leaf. The dataset contains three classes, two disease classes and one healthy potato class. Models are trained by different train-test splits to understand better and get accurate results. To test performance of the data Applied Accuracy Precision, Recall, F1 score and ROC/AUC curve are used. By using Gradient Boosting approach results are better even for mostly effected leaf.