CODE-MNET: Coconut Leaf Disease Classification Using Deep Learning Based MobileNet

Coconut is considered a “functional food” because of its rich nutritional content, which includes healthy fats, vitamins, and minerals. It provides essential nutrients such as fiber, potassium, and magnesium that contribute to overall health. Coconut leaf disease (CLD) poses a significant threat to coconut production worldwide, requiring efficient and accurate detection methods. To overcome these challenges, a novel deep learning-based CODE-MNET model has been proposed for classifying the CLD. Initially, the dataset was gathered from coconut farms in the Chennai region of Tamil Nadu and the collected image of a coconut leaf is taken as a test image to analyse whether it is healthy or diseased. The gathered images are pre-processed using a gaussian filter for eliminating the noise artifacts and adjusting the image size to make it uniform for analysis. To increase the diversity of training data by utilizing random transformations such as rotating, flipping to generate additional variations of the original image. An important color feature can be extracted from the pre-processed image by using MobileNet. Then, the extracted features are fed into various machine learning (ML) algorithms such as support vector machine (SVM), random forest (RF), decision tree (DT) and K-Nearest Neighbors (KNN) to classify the image as either healthy or diseased. The classification accuracy of 98.81% for the proposed SVM is highly reliable for publicly available dataset. The proposed CODE-MNET model achieves the overall accuracy by 5.15%, 19.03% and 16.95% compared to the existing methods such as CNN, YOLOv9 and Federated learning.