Mango Leaf Disease Identification Using Deep Learning

Abstract

Mango (Mangifera indica) is the most economically important fruit crop in Bangladesh but production of mango is hampered by several leaf diseases, which greatly reduce its successful growth and concomitant yield. Early and accurate diagnosis of these diseases is crucial for the livelihoods of farmers, to prevent crop loss and ascribe to sustainable farming. Conventional hand scouting procedures are labor intensive, subjective, and not readily available to the smallholder farmer. In these contexts, this work proposes a deep-learning-based automatic method for recognizing and diagnosing six prominent mango leaf diseases: Anthracnose, Bacterial Canker, Gall Midge, Healthy., Powdery Mildew and Sooty Mould. The dataset contains a collection of 2,364 wine images acquired from May 2025 to August 2025 by an iPhone 11 in HTEC format which was later transferred into JPG for preprocessing. Following extensive image processing, involving resizing, background removal, gree masking, morphological filtering, segmentation and severity scoring the dataset was enlarged to 6000 balanced images (1000 per class). These features were separated into 70%Training, 20%Validation and 10%Testing image sets.