Apple Leaf Disease Detection Using Transfer Learning

Abstract

With the increasing demand for sustainable and efficient agricultural practices, precision agriculture has emerged as a key paradigm for optimizing crop management, the role of precision farming is becoming increasingly vital. This is particularly significant in the horticulture sector, which accounts for about 30% of India's agricultural GDP, with apple cultivation being a key component. This paper emphasizes the use of sophisticated deep learning techniques, specifically for detecting and classifying diseases in apple leaves, a vital factor in maintaining crop health and optimizing yields. In this paper we have proposed a transfer learning based model TransferNet, utilizing RESNET50V2 as the base model with weights initialized from ImageNet and additional layers incorporated for the analysis of apple leaf images, using data from the FGCV-8 dataset of the Plant Pathology 2021 Kaggle Competition. The paper goes beyond merely achieving high accuracy in disease detection. It also addresses the real-world challenges such as varying light conditions, obstructions, and the different orientations of leaves. The study's results surpasses its predecessor models, delivering F1-score of 91.03%, accuracy of 91.63%, a precision of 94.14% and a recall of 88.12% on the FGCV-8 dataset. This approach, leveraging deep learning, not only marks an advancement in disease detection technology but also signifies a crucial step forward in the development of intelligent agricultural systems.