Deep Learning and Computer Vision based Model for Detection of Diseased Mango Leaves

Abstract

Mangifera Indica, commonly known as mangoes, is the most commercialized export fruit crop in India, accounting for about 40% of the total global production. Due to its widespread production, it is vulnerable to a variety of diseases that affect its yield and resulting in loss. These diseases like Anthracnose, Powdery Mildew, Leaf blights, etc., occur primarily on leaves. As a result, there is a great need for a system that helps in the detection of diseased mango leaves. In this paper, we propose a system that makes use of pre-trained Convolutional Neural Network architecture, the ResNet-50 for the detection of infected mango leaves. The dataset contains 435 images of mango leaves with binary classification as healthy and diseased. These images are pre-processed by resizing them and applying CLAHE. After applying in-place data augmentation on the dataset, the features are extracted using the ResNet-50 model. For the classification process, we make use of fine-tuned head and Machine Learning classifiers such as Support Vector Machine, Gradient Boosting, Logistic Regression, XGBoost, Decision Tree, and K Nearest Neighbour. Among them, the fine-tuned head classifier achieved an accuracy of 97.7%, and Machine Learning classifiers such as SVM, Logistic Regression obtained an accuracy of 100%. The experimental results obtained validate that the system is efficient in its performance of detecting the two classes of mango leaves.