EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A Dual Branch Fusion Network for Simultaneous Tea Leaf Disease Diagnosis and Age-Based Quality Grade Evaluation

Xin Zhang iD, Jiahua Ren iD, Siyu Qin, Xiu Zhang

Plants · 04.08.2026 · 15 (15) · 2391

DOI: 10.3390/plants15152391 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Plants, online erschienen am 04.08.2026. Verfasst von Xin Zhang und 3 weiteren Autor:innen. Beteiligte Einrichtung: Interdisciplinary Center, Tianjin Normal University, Tianjin 300387, China. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The simultaneous diagnosis of diseases and evaluation of age quality grades in tea leaves are critical for precision agriculture and the economic valuation of tea products. Although deep learning has shown promise in agricultural vision tasks, current multi-task models often suffer from performance degradation due to feature conflicts: tea leaf disease recognition relies heavily on macro-structural lesions, whereas tea leaf-age quality grading depends on micro-textural features such as trichome density and color uniformity. To address this discrepancy, we propose a novel dual branch fusion network. Our architecture fundamentally decouples the feature extraction process by utilizing a dual branch mechanism. The first branch employs global average pooling to capture first-order spatial statistics; it can retain the global structural layout necessary for macro-lesion detection. The second branch introduces a dimensionality-reduced self-bilinear pooling module to compute second-order covariance matrices; it can effectively capture the fine-grained textural patterns essential for micro-grade classification. These decoupled features are subsequently fused and optimized through a weighted multi-task loss function. Experimental results on a comprehensive tea leaf dataset demonstrate that the proposed dual fusion framework significantly outperforms baseline models. The proposed network can rescue the disease classification accuracy drop observed in standard bilinear models while maintaining exceptional grading performance. Furthermore, the proposed network maintains a compact parameter footprint and low computational complexity. This balance renders it suitable for deployment on agricultural Internet of Things edge devices where inference speed is critical.

Schlagwörter

Stichwörter der Autor:innen

deep learningmulti-task learningprecision agriculturetea leaf disease and pest-damagetea leaf-age quality grading

Autor:innen und Institutionen

  • Interdisciplinary Center, Tianjin Normal University, Tianjin 300387, China.
  • Tianjin Key Laboratory of Wireless Mobile Communications and Power Transmission, Tianjin Normal University, Tianjin 300387, China.
  • College of Artificial Intelligence, Tianjin Normal University, Tianjin 300387, China.

ORCID-Profile: Xin Zhang, Jiahua Ren

Zitieren

APA

Zhang, X., Ren, J., Qin, S., & Zhang, X. (2026). A Dual Branch Fusion Network for Simultaneous Tea Leaf Disease Diagnosis and Age-Based Quality Grade Evaluation. Plants, 15(15), 2391. https://doi.org/10.3390/plants15152391

Vancouver

Zhang X, Ren J, Qin S, Zhang X. A Dual Branch Fusion Network for Simultaneous Tea Leaf Disease Diagnosis and Age-Based Quality Grade Evaluation. Plants. 2026;15(15):2391. doi:10.3390/plants15152391

BibTeX
@article{Zhang2026,
  author = {Zhang, Xin and Ren, Jiahua and Qin, Siyu and Zhang, Xiu},
  title = {A Dual Branch Fusion Network for Simultaneous Tea Leaf Disease Diagnosis and Age-Based Quality Grade Evaluation},
  journal = {Plants},
  year = {2026},
  volume = {15},
  number = {15},
  pages = {2391},
  doi = {10.3390/plants15152391}
}
Kostenlos starten →