Eggplant Leaf Disease Classification Using Deep Learning and a Robust Real-Time System Based on YOLOv8 and Streamlit

Abstract

Early disease Classification will help reduce crop loss as well as increase agricultural productivity. A rapid and accurate deep learning-based framework to identify different diseases in eggplant on the marketplace level has been proposed in the research. Implementation and Training of an End-to-End Object Classification Based on YOLOv8. To train a custom multi-class data set, six target classes: Healthy Leaf, White Mold Disease, Leaf Spot Disease, Wilt Disease, Mosaic Virus Disease, and Insect Pest Disease. Initial work was done in developing the object detectors. We used standard metrics like accuracy, precision, recall, and F1-score to evaluate the performance of the model. When it came to finding Plant Leaf Disease (PLD), traceable configurations were revealed from configuration tuning among the combinations of hyperparameters, which converged at equal measurement intervals on a curve between Classification accuracy and computation efficiency from screened candidate architectures along ranges determined by performance metrics defined for detecting plant diseases using only above-ground debris as input sources. With stable convergence during the training phase, the model YOLOv8m had an accuracy of 96.84%, a precision of 96.85%, a recall of 96.84%, and an F1 score of 96.84%. The model that has been trained was deployed with the help of a web application named Streamlit, so that it could be used in practical procedures where disease can be detected if we upload an image. That means the system is robust and operates effectively in the wild as opposed to ideal test conditions, which lends itself well to agricultural usage. The present work provides an integrated, optimized deep learning detector with a user interface beneficial for precision farming that can help in the early identification of diseases on eggplants, resulting in increased yield.