Mulberry leaf classification based on a hybrid deep learning approach - DCGAN synthesis and EfficientNetB0
International Journal of Drug Delivery Technology · 08.06.2026 · 16 (50s)
DOI: 10.25258/ijddt.16.50s.76 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Drug Delivery Technology, online erschienen am 08.06.2026. Verfasst von Hemanth Chandra N und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Sericulture is an important part of agriculture. The productivity, both in terms of quality and quantity of silk in sericulture, relies on the cultivation of Mulberry leaves, as silkworms’ main source of food is mulberry leaves. Silk worms fed on healthy leaves produce high-quality silk. Productivity and nutrient value of Mulberry leaves often suffer from diseases that appear seasonally or regularly, like leaf rust, leaf spot, mildew, etc. These diseases spread rapidly, reducing the nutrient value of leaves, which leads to a reduction in silk production. Manual detection of diseased leaves is timeconsuming and often inaccurate; therefore, using deep learning technology for disease identification and leaf classification is more effective and faster. Hence, in this paper we suggest that CNN-based EfficientNetB0 model used for classification; EfficientNetB0 is baseline models of the EfficientNet family that achieve high accuracy with fewer parameters and lower computational cost compared to traditional CNNs, It can work on a wide range of datasets, but it’s good to have at least a few thousand images per class to avoid overfitting for good performance. Hence, in this research, we use Generative Adversarial Networks (GANs) for high-fidelity synthetic image generation that helps to expand the original dataset. The original dataset, which contains 1091 real images belonging to 3 different classes, was enlarged using Deep Convolutional GAN (DCGAN) to over 2000 images in each class, resulting in a balanced dataset of 6000 images. The dataset split ratio followed is 70:15:15 (training: validation: test). Lightweight tensor flow-based model EfficientNetB0 was fine-tuned and trained on the enhanced dataset and achieved an accuracy of 99%. Results demonstrate the effectiveness of the EfficientNetB0 with a limited size of dataset.
Zitieren
APA
N, H. C., Vidyashree, V., & J, T. (2026). Mulberry leaf classification based on a hybrid deep learning approach - DCGAN synthesis and EfficientNetB0. International Journal of Drug Delivery Technology, 16(50s). https://doi.org/10.25258/ijddt.16.50s.76
Vancouver
N HC, Vidyashree V, J T. Mulberry leaf classification based on a hybrid deep learning approach - DCGAN synthesis and EfficientNetB0. International Journal of Drug Delivery Technology. 2026;16(50s). doi:10.25258/ijddt.16.50s.76
BibTeX
@article{N2026,
author = {N, Hemanth Chandra and Vidyashree, Vidyashree and J, Thriveni},
title = {Mulberry leaf classification based on a hybrid deep learning approach - DCGAN synthesis and EfficientNetB0},
journal = {International Journal of Drug Delivery Technology},
year = {2026},
volume = {16},
number = {50s},
doi = {10.25258/ijddt.16.50s.76}
}
EUVIMED