Article Per Year (5 Year)
p-Index From 2020 - 2025
0.23
P-Index
This Author published in this journals
All Journal Journal of Scientific Research, Education, and Technology
Utomo, Ragil Gigih
Unknown Affiliation
Computer Science & IT Economics, Econometrics & Finance Education Engineering Social Sciences

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

 1 
Enhancing Convolutional Neural Network Accuracy for Herbal Leaf Classification Using Squeeze and Excitation Attention Utomo, Ragil Gigih; Mardhiyyah, Rodhiyah
Journal of Scientific Research, Education, and Technology (JSRET) Vol. 4 No. 4 (2025): Vol. 4 No. 4 2025
Publisher : Kirana Publisher (KNPub)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.58526/jsret.v4i4.938

Abstract

Accurate identification of herbal plant species is crucial for human health, but remains challenging, even for experts. This study addressed this need by developing a Convolutional Neural Network (CNN) model integrated with an attention mechanism for reliable herbal leaf classification. The research implemented the MobileNetV2 architecture, which was enhanced by incorporating the Squeeze-and-Excitation (SE) attention module. The dataset consisted of 1,500 images across 10 classes of herbal leaves, split into 80% for training, 10% for validation, and 10% for testing. Both the native CNN and the enhanced CNN (CNN-AM) models were trained using TensorFlow and evaluated using standard metrics like accuracy, precision, recall, and F1-score. The comparison results decisively demonstrated the effectiveness of the attention mechanism. Integrating Squeeze-and-Excitation significantly improved performance. The average accuracy of the model increased from 68% to 72%, while the average loss decreased from 1.03 to 1.02. The best-performing CNN-AM model achieved a strong 86% accuracy with a 0.53 loss. These findings confirm that the Squeeze-and-Excitation attention mechanism effectively enhances herbal leaf classification performance, offering a promising foundation for developing reliable and efficient identification systems.
Co-Authors Rodhiyah Mardhiyyah