EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Rock Melon Ripeness Detection Using YOLOv8 and SE-Lightweight YOLOv8

Dewi Kania Widyawati, Oki Arifin, Sylvia, Zuriati, Fahri Ali, Sela Wissi Yani, Dede Aprizal

Engineering, Technology & Applied Science Research · 04.04.2026 · 16 (2) · 33518-33524

DOI: 10.48084/etasr.16236 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Engineering, Technology & Applied Science Research, online erschienen am 04.04.2026. Verfasst von Dewi Kania Widyawati und 6 weiteren Autor:innen. Die Arbeit nennt 22 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate detection of fruit ripeness is crucial for improving harvest efficiency and supporting smart agriculture systems. This study investigates the performance of You Only Look Once version 8 (YOLOv8) and Squeeze-and-Excitation Lightweight YOLOv8 (SE-Lightweight YOLOv8) for automatic rock melon ripeness detection based on visual features. A dataset of 2,000 high-resolution rock melon images was collected under real field conditions and annotated into 2,014 rock melon objects, categorized into three ripeness levels: fully ripe, half-ripened, and unripe. Model performance was evaluated using precision, recall, accuracy, F1-score, and mean Average Precision (mAP). Experimental results indicate that both models achieve comparably high detection performance. YOLOv8 demonstrates slightly higher precision (0.994), F1-score (0.994), and overall accuracy (0.992), reflecting a more balanced detection capability. In contrast, SE-Lightweight YOLOv8 attains a marginally higher recall (0.999), indicating improved detection of relevant melon ripeness instances. In terms of localization performance, both models achieve similar mAP50–95 scores (0.925) and nearly identical mAP50 values (0.995 for YOLOv8 and 0.994 for SE-Lightweight YOLOv8), suggesting comparable robustness across different Intersection-Over-Union (IoU) thresholds. These findings highlight a trade-off between detection accuracy and computational efficiency. YOLOv8 is better suited for applications that require high detection accuracy, whereas SE-Lightweight YOLOv8 is a viable alternative for real-time deployment on resource-constrained devices. The results demonstrate the potential of YOLO-based object detection models to support precision agriculture and optimize rock melon harvesting processes.

Zitieren

APA

Widyawati, D. K., Arifin, O., Sylvia, Zuriati, Ali, F., Yani, S. W., & Aprizal, D. (2026). Rock Melon Ripeness Detection Using YOLOv8 and SE-Lightweight YOLOv8. Engineering, Technology & Applied Science Research, 16(2), 33518-33524. https://doi.org/10.48084/etasr.16236

Vancouver

Widyawati DK, Arifin O, Sylvia, Zuriati, Ali F, Yani SW, et al. Rock Melon Ripeness Detection Using YOLOv8 and SE-Lightweight YOLOv8. Engineering, Technology & Applied Science Research. 2026;16(2):33518-33524. doi:10.48084/etasr.16236

BibTeX
@article{Widyawati2026,
  author = {Widyawati, Dewi Kania and Arifin, Oki and Sylvia and Zuriati and Ali, Fahri and Yani, Sela Wissi and Aprizal, Dede},
  title = {Rock Melon Ripeness Detection Using YOLOv8 and SE-Lightweight YOLOv8},
  journal = {Engineering, Technology & Applied Science Research},
  year = {2026},
  volume = {16},
  number = {2},
  pages = {33518--33524},
  doi = {10.48084/etasr.16236}
}
Kostenlos starten →