YOLOv11-HRS: An Improved Model for Strawberry Ripeness Detection DOI Creative Commons
Jianhua Liu, Jing Guo, Shouchuan Zhang

et al.

Agronomy, Journal Year: 2025, Volume and Issue: 15(5), P. 1026 - 1026

Published: April 25, 2025

Automated ripeness detection in large-scale strawberry cultivation is often challenged by complex backgrounds, significant target scale variation, and small object size. To address these problems, an efficient model, YOLOv11-HRS, proposed. This model incorporates a hybrid channel–space attention mechanism to enhance its key features reduce interference from backgrounds. Furthermore, the RepNCSPELAN4_L module devised multi-scale representation through contextual feature aggregation. Simultaneously, 160 × small-target head embedded pyramid capability of targets. It replaces original SPPF with higher-performance SPPELAN further accuracy. Experimental results on self-constructed dataset SRD show that YOLOv11-HRS improves [email protected] [email protected]:0.95 3.4% 6.3%, respectively, reduces number parameters 19%, maintains stable inference speed compared baseline YOLOv11 model. study presents practical solution for natural environments. also provides essential technical support advancing intelligent management cultivation.

Language: Английский

LRNTRM-YOLO: Research on Real-Time Recognition of Non-Tobacco-Related Materials DOI Creative Commons
Chunjie Zhang, Lijun Yun,

Chenggui Yang

et al.

Agronomy, Journal Year: 2025, Volume and Issue: 15(2), P. 489 - 489

Published: Feb. 18, 2025

The presence of non-tobacco-related materials can significantly compromise the quality tobacco. To accurately detect materials, this study introduces a lightweight and real-time detection model derived from YOLOv11 framework, named LRNTRM-YOLO. Initially, due to sub-optimal accuracy in detecting diminutive was augmented by incorporating an additional layer dedicated enhancing small targets, thereby improving overall accuracy. Furthermore, attention mechanism incorporated into backbone network focus on features efficacy model. Simultaneously, for introduction SIoU loss function, angular vector between bounding box regressions utilized define thus training efficiency Following these enhancements, channel pruning technique employed streamline network, which not only reduced parameter count but also expedited inference process, yielding more compact material detection. experimental results NTRM dataset indicate that LRNTRM-YOLO achieved mean average precision (mAP) 92.9%, surpassing baseline margin 4.8%. Additionally, there 68.3% reduction parameters 15.9% decrease floating-point operations compared Comparative analysis with prominent models confirmed superiority proposed terms its architecture, high accuracy, capabilities, offering innovative practical solution future.

Language: Английский

Citations

1
AI-Driven Irrigation Systems for Sustainable Water Management: A Systematic Review and Meta-Analytical Insights DOI Creative Commons
Gülcay ERCAN OĞUZTÜRK

Smart Agricultural Technology, Journal Year: 2025, Volume and Issue: unknown, P. 100982 - 100982

Published: May 1, 2025

Language: Английский

Citations

0
YOLOv11-HRS: An Improved Model for Strawberry Ripeness Detection DOI Creative Commons
Jianhua Liu, Jing Guo, Shouchuan Zhang

et al.

Agronomy, Journal Year: 2025, Volume and Issue: 15(5), P. 1026 - 1026

Published: April 25, 2025

Automated ripeness detection in large-scale strawberry cultivation is often challenged by complex backgrounds, significant target scale variation, and small object size. To address these problems, an efficient model, YOLOv11-HRS, proposed. This model incorporates a hybrid channel–space attention mechanism to enhance its key features reduce interference from backgrounds. Furthermore, the RepNCSPELAN4_L module devised multi-scale representation through contextual feature aggregation. Simultaneously, 160 × small-target head embedded pyramid capability of targets. It replaces original SPPF with higher-performance SPPELAN further accuracy. Experimental results on self-constructed dataset SRD show that YOLOv11-HRS improves [email protected] [email protected]:0.95 3.4% 6.3%, respectively, reduces number parameters 19%, maintains stable inference speed compared baseline YOLOv11 model. study presents practical solution for natural environments. also provides essential technical support advancing intelligent management cultivation.

Language: Английский

Citations

0