Custom CNN Model for Mango Leaf Disease Detection

Abstract

Plant health disorders can affect the productivity of crops adversely, so it's important to find them early in farming. This research develops a deep learning-based system for the identification of mango leaf diseases utilizing image data. We built and trained a custom Convolutional Neural Network (CNN) from scratch on the Mango Leaf BD dataset, which has eight types of healthy and diseased leaves. For making the images more generalized, they have gone under the procedure of resizing and normalizing before the data augmentation techniques are used. Standard evaluation criteria like accuracy, precision, recall, and F1-score are utilized to test the model, and it does well on the test dataset. Also, a desktop-based graphical user interface (GUI) is made with Python and Tkinter, which makes it easy to make predictions for one image or a group of images. The system works completely offline, so it can be used in places with few resources. It can also be expanded for use in real-world farming situations.