Taufik Hidayah, Muhammad (2025) Klasifikasi Jenis Penyakit Pada Daun Mangga Menggunakan Metode Convolutional Neural Network. Other thesis, Universitas Islam Riau.

Text 203510507.pdf - Submitted Version Restricted to Registered users only Download (7MB) | Request a copy

Abstract

Mango plants (Mangifera indica) are one of the important horticultural commodities in Indonesia that have high economic value. In addition to being a source of income for farmers, mangoes are also widely consumed by the community, both fresh and processed. However, the productivity of mango plants often decreases due to attacks by various types of diseases on their leaves, which can inhibit the process of photosynthesis and plant growth. Therefore, early identification of mango leaf diseases is very important to improve the quality and quantity of the harvest. In this study, the convolutional neural network (CNN) method was used to classify mango plant diseases through leaf images using CNN architectures, namely MobileNetV2, Vgg16, and Xception. The dataset used was obtained from the Kaggle website and combined with real data in the field, with a total of 2825 images consisting of 5 disease classes, namely Anthracnose, Bacterial Canker, Die Back, Gall Midge, and healthy leaves. The dataset was trained with augmentation to increase data variation, then each architecture was tested with 8 parameter selection scenarios, including batch size, epoch, and learning rate. The test results showed that the model with the best accuracy on the MobileNetV2 architecture was found in the 7th scenario with an accuracy of 97.87%. The VGG16 architecture produced the best model in the 3rd scenario with an accuracy of 96.10%, while the Xception architecture produced the best model in the 1st scenario with an accuracy of 97.16%. The implementation of this research was carried out based on a website using the Flask framework, where predictions from each architecture with the best model were integrated, and the final prediction of the disease was taken based on the highest accuracy

Actions (login required)

View Item