EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A Novel CGT‐YOLO Algorithm for Silkworm Health State Detection in Diverse Environments

Peng Chen, Tin Tin Ting, Siti Sarah Maidin, Fang Wan, Tiankai Wei, Caijun Gui, Chuqi Ma

IET Computer Vision · 2026-01 · 20 (1)

DOI: 10.1049/cvi2.70096 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IET Computer Vision, online erschienen am 16.11.2026. Verfasst von Peng Chen und 6 weiteren Autor:innen. Die Arbeit nennt 48 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Automated silkworm health‐state detection faces distinct challenges in real farming environments, including dense small‐target distribution, visually similar rearing backgrounds, and variable imaging conditions. To address these challenges, this paper proposes CGT‐YOLO, a three‐stage object detection algorithm built upon YOLOv8n for binary classification of healthy and diseased silkworm bodies. Three complementary modules are introduced: FormNet replaces the standard C2f backbone by integrating ConvFormer Block_CGLU and Transformer Block_CGLU to enhance small‐object feature extraction while maintaining computational efficiency; C2fGSBneck refines the neck network for more precise multi‐scale feature fusion under complex backgrounds; and Detect_TaskNet introduces a redesigned detection head with integrated dynamic feature alignment, improving recognition accuracy and robustness across varied conditions. Ablation experiments on the validation set demonstrate that CGT‐YOLO achieves 93.5% precision, 88.6% recall, and 95.6% mAP@0.5, surpassing the YOLOv8n baseline by 7.3, 11.3, and 9.4 percentage points respectively. Comparative experiments on the test set further show that CGT‐YOLO achieves the best detection performance among all compared methods, with 94.8% precision, 89.3% recall, and 95.8% mAP@0.5, providing a practical and efficient solution for real‐time silkworm health‐state monitoring in diverse farming environments. Additional validation on the Corn Weed and CrowdHuman datasets, with models independently retrained on each dataset, demonstrates that CGT‐YOLO maintains consistent performance advantages over the YOLOv8n baseline across diverse object detection tasks, supporting its potential for broader agricultural deployment.

Zitieren

APA

Chen, P., Ting, T. T., Maidin, S. S., Wan, F., Wei, T., Gui, C., & Ma, C. (2026). A Novel CGT‐YOLO Algorithm for Silkworm Health State Detection in Diverse Environments. IET Computer Vision, 20(1). https://doi.org/10.1049/cvi2.70096

Vancouver

Chen P, Ting TT, Maidin SS, Wan F, Wei T, Gui C, et al. A Novel CGT‐YOLO Algorithm for Silkworm Health State Detection in Diverse Environments. IET Computer Vision. 2026;20(1). doi:10.1049/cvi2.70096

BibTeX
@article{Chen2026,
  author = {Chen, Peng and Ting, Tin Tin and Maidin, Siti Sarah and Wan, Fang and Wei, Tiankai and Gui, Caijun and Ma, Chuqi},
  title = {A Novel CGT‐YOLO Algorithm for Silkworm Health State Detection in Diverse Environments},
  journal = {IET Computer Vision},
  year = {2026},
  volume = {20},
  number = {1},
  doi = {10.1049/cvi2.70096}
}
Kostenlos starten →