Deep Learning-Based Mango Leaf Disease Classification Using Convolutional Neural Networks

Abstract

Abstract - Mango (Mangifera indica) is one of the most commercially significant and nutritious tropical fruit crops. However, yield and fruit quality are significantly reduced by several leaf diseases, such as powdery mildew, sooty mold, anthracnose, bacterial canker, and gall midge infestation. Traditional method of visually diagnosing these diseases is laborious, prone to mistakes, time consuming, and often unavailable to many small holder farmers, this research used deep learning-based approaches that use convolutional neural networks (CNN) to classify mango leaf diseases. In order to evaluate this research method, a dataset of 4000 images were taken over eight classes, including a class representing healthy and diseased leaves. The images in the dataset were also increased to help improve the CNN’s ability to generalize from a limited set of images. Using TensorFlow, the EfficientNetB0 architecture with pre-trained weights, achieved a validation accuracy of approximately 98% and produced high precision, recall, and f1-scores throughout testing. In order to make the CNN model practical to apply, it was converted to TensorFlow Lite and then incorporated into a mobile application developed using Flutter, which enables real time classification of diseases on mango leaves. The proposed system demonstrates the potential of CNN-based models to provide accessible, scalable and field-ready solutions for early detection of mango diseases, ultimately leading to better crop management and productivity. Keywords: Mango disease detection, deep learning, Convolutional Neural Networks, image classification, EfficientNet, mobile deployment.

Plant phenotyping relevance

マンゴー葉の画像から病害状態を分類するCNN手法の開発・評価が中心であり、植物の病徴を直接推定するため、植物フェノタイピング手法として採用。

Code and data availability

公開本文の所在を確認できませんでした。非公開または購読が必要な可能性があります。

No evidence-backed public reproduction asset is currently recorded.