EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

Mushroom Species Image Classification Using EfficientNetB4 with MixUp and CutMix Augmentation

Sike Laurika Olia

Devotion : Journal of Research and Community Service · 25.10.2026 · 7 (10) · 1221-1239

DOI: 10.59188/devotion.v7i10.25789 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Devotion : Journal of Research and Community Service, online erschienen am 25.10.2026. Verfasst von Sike Laurika Olia. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Mushroom image classification is an application of computer vision that aims to recognize mushroom categories from the visual characteristics of an image.

Objective

This study evaluates EfficientNetB4 with MixUp and CutMix augmentation for 21-class mushroom image classification.

Method

The initial dataset contained 8,857 images and 20,248 COCO annotations. Following bounding-box cropping and the removal of 107 crops that were too small, 20,141 crops were retained and split into 14,289 training, 2,461 validation, and 3,391 testing images. Four configurations were compared: a baseline, MixUp (α = 0.2), CutMix (α = 1.0), and a combination of MixUp and CutMix. The model used ImageNet-pretrained EfficientNetB4 with 380 × 380 inputs, five feature-extraction epochs, and 15 fine-tuning epochs.

Result

The combined configuration achieved the highest validation accuracy (88.30%), whereas CutMix achieved the highest validation macro F1-score (88.08%). CutMix was selected using the macro F1-score and achieved 86.17% testing accuracy, 84.51% macro precision, 86.60% macro recall, 85.08% macro F1-score, and 95.72% Top-3 accuracy.

Conclusion

CutMix provided slightly better class-level balance on validation, but errors remained among visually similar species, and the 90% accuracy target was not achieved. The study therefore does not demonstrate a clear benefit of MixUp or CutMix over the baseline.

Zitieren

APA

Olia, S. L. (2026). Mushroom Species Image Classification Using EfficientNetB4 with MixUp and CutMix Augmentation. Devotion : Journal of Research and Community Service, 7(10), 1221-1239. https://doi.org/10.59188/devotion.v7i10.25789

Vancouver

Olia SL. Mushroom Species Image Classification Using EfficientNetB4 with MixUp and CutMix Augmentation. Devotion : Journal of Research and Community Service. 2026;7(10):1221-1239. doi:10.59188/devotion.v7i10.25789

BibTeX
@article{Olia2026,
  author = {Olia, Sike Laurika},
  title = {Mushroom Species Image Classification Using EfficientNetB4 with MixUp and CutMix Augmentation},
  journal = {Devotion : Journal of Research and Community Service},
  year = {2026},
  volume = {7},
  number = {10},
  pages = {1221--1239},
  doi = {10.59188/devotion.v7i10.25789}
}
Kostenlos starten →