EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

Comparative Analysis of CNN and YOLO for Aromatic Leaf Detection on Android-Based Deep Learning Applications

Yuli Safrina, Muhammad Fikry, Mukhlis Abd Muthalib

Journal of Applied Informatics and Computing · 12.06.2026 · 10 (3) · 2489-2501

DOI: 10.30871/jaic.v10i3.13015 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Applied Informatics and Computing, online erschienen am 12.06.2026. Verfasst von Yuli Safrina und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Commonly used aromatic leaves in Indonesian cuisine include bay leaves (Syzygium polyanthum), pandan leaves (Pandanus amaryllifolius), lime leaves (Citrus hystrix), curry leaves (Murraya koenigii), and turmeric leaves (Curcuma longa). Their similar shapes, colors, and textures often make manual identification difficult. Therefore, deep learning technology can be utilized to automatically identify and detect aromatic leaf types through digital images. This study aims to analyze the performance of a Convolutional Neural Network (CNN) using the EfficientNet-B0 architecture and the YOLOv11 model with the AdamW optimizer in detecting and classifying aromatic leaves. The system is implemented using a Python Flask framework for the web based backend and Flutter for the mobile application interface on Android devices. The dataset used in this study consists of 671 digital images obtained through direct image collection and supporting datasets. The dataset is categorized into five classes: bay leaf, pandan leaf, lime leaf, curry leaf, and turmeric leaf. Furthermore, the dataset is divided into training data (89%), validation data (7%), and testing data (4%). The results show that the YOLOv11 model outperforms the CNN (EfficientNet-B0) model. YOLOv11 achieved a precision of 73.36%, recall of 84.35%, mAP50 of 83.93%, and mAP50-95 of 71.38%. Meanwhile, EfficientNet-B0 achieved a best validation accuracy of 81.40% and a test accuracy of 62.07%. Based on experimental results, YOLOv11 demonstrates higher detection confidence and more consistent performance compared to EfficientNet-B0. In addition, YOLOv11 is more suitable for mobile deployment due to its real-time object detection capability with faster inference speed, while the system is supported by a Flask based backend and a Flutter mobile application interface.

Zitieren

APA

Safrina, Y., Fikry, M., & Muthalib, M. A. (2026). Comparative Analysis of CNN and YOLO for Aromatic Leaf Detection on Android-Based Deep Learning Applications. Journal of Applied Informatics and Computing, 10(3), 2489-2501. https://doi.org/10.30871/jaic.v10i3.13015

Vancouver

Safrina Y, Fikry M, Muthalib MA. Comparative Analysis of CNN and YOLO for Aromatic Leaf Detection on Android-Based Deep Learning Applications. Journal of Applied Informatics and Computing. 2026;10(3):2489-2501. doi:10.30871/jaic.v10i3.13015

BibTeX
@article{Safrina2026,
  author = {Safrina, Yuli and Fikry, Muhammad and Muthalib, Mukhlis Abd},
  title = {Comparative Analysis of CNN and YOLO for Aromatic Leaf Detection on Android-Based Deep Learning Applications},
  journal = {Journal of Applied Informatics and Computing},
  year = {2026},
  volume = {10},
  number = {3},
  pages = {2489--2501},
  doi = {10.30871/jaic.v10i3.13015}
}
Kostenlos starten →