EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

Improving Accuracy in Deep Learning-Based Mushroom Image Classification through Optimal Use of Classification Techniques

Johan Muliadi Kerta, Abdul Haris Rangkuti, Sian Lun Lau, Albert Kurniawan, Melanie Gabriela, Alicia Tandianto

JOIV : International Journal on Informatics Visualization · 31.05.2025 · 9 (3) · 1298

DOI: 10.62527/joiv.9.3.2820 ↗

Kurz gefasst

Fachartikel in der Zeitschrift JOIV : International Journal on Informatics Visualization, online erschienen am 31.05.2025. Verfasst von Johan Muliadi Kerta und 5 weiteren Autor:innen. Die Arbeit wurde bisher 5-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The primary purpose of this research is to address the existing knowledge gap surrounding various lesser-known types of edible mushrooms. A common understanding exists that mushrooms are edible and possess numerous health benefits. This research is intended to advance that understanding by deploying AI technology and deep learning models specifically designed to recognize and identify various fungi. During this research, we have developed a unique derivative of deep learning. This involved testing several Convolutional Neural Network (CNN) models aimed at automatically identifying and detecting different types of mushrooms and understanding the benefits associated with each type. The research methodology was divided into several stages: Collection of mushroom images, Preprocessing of images, Feature extraction, and Classification. The preprocessing involved adjustments such as scale, image rotation, and setting the brightness range. The goal of selecting and training the CNN model was to enhance the classification accuracy of mushroom images within each class. The data was divided into training, testing, and validation sets for the experimental stage. The purpose was to process image data from test and validation images based on the training images that have been processed. We evaluated the classification layer to be shorter, but it demonstrated excellent accuracy in assessing similarity performance. Based on several experiments conducted using different CNN models, DenseNet, MobileNetV2, and InceptionResNetV2 models achieved an accuracy of more than 90%, specifically 95%, 94%, and 92%, respectively. The most accurately recognized mushroom types include Snow, Dried Shitake, King Oyster, Straw, Button, and Truffle; some CNN models could identify these up to 100%. Overall, the models and algorithms used in this research successfully facilitated the identification and detection of various types of fungi. They were fast and displayed high accuracy performance. Hopefully, this research can be extended to process images of even more diverse types of mushrooms, particularly in terms of shape, color, and texture characteristics. This will enhance the depth and breadth of knowledge in this field and further advance our understanding of the beneficial properties of various mushrooms.

Zitieren

APA

Kerta, J. M., Rangkuti, A. H., Lau, S. L., Kurniawan, A., Gabriela, M., & Tandianto, A. (2025). Improving Accuracy in Deep Learning-Based Mushroom Image Classification through Optimal Use of Classification Techniques. JOIV : International Journal on Informatics Visualization, 9(3), 1298. https://doi.org/10.62527/joiv.9.3.2820

Vancouver

Kerta JM, Rangkuti AH, Lau SL, Kurniawan A, Gabriela M, Tandianto A. Improving Accuracy in Deep Learning-Based Mushroom Image Classification through Optimal Use of Classification Techniques. JOIV : International Journal on Informatics Visualization. 2025;9(3):1298. doi:10.62527/joiv.9.3.2820

BibTeX
@article{Kerta2025,
  author = {Kerta, Johan Muliadi and Rangkuti, Abdul Haris and Lau, Sian Lun and Kurniawan, Albert and Gabriela, Melanie and Tandianto, Alicia},
  title = {Improving Accuracy in Deep Learning-Based Mushroom Image Classification through Optimal Use of Classification Techniques},
  journal = {JOIV : International Journal on Informatics Visualization},
  year = {2025},
  volume = {9},
  number = {3},
  pages = {1298},
  doi = {10.62527/joiv.9.3.2820}
}
Kostenlos starten →