Lightweight Salix Cheilophila Recognition Method Based on Improved YOLOv8n DOI
Haotian Ma, Zhigang Liu,

C. C. Pei

et al.

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 11, 2024

Abstract Stumping is an important measure for the care and management of salix cheilophila during its growth. Rapid accurate detection in stumping period desert basis intelligent equipment. However, complex model needs high computing power hardware. It limits deployment application recognition Therefore, this study took areas Shierliancheng, Inner Mongolia Autonomous Region as research object, proposed improved YOLOv8 rapid identification method, named YOLOV8-VCAD. First, lightweight network VanillaNet was used to replace backbone lessen load complexity model. Coordinate attention mechanism embedded extract features by setting location information, which strengthened regression positioning abilities Second, introducing adaptive feature fusion pyramid significantly strengthens model's ability characterize integrate features, improving accuracy performance target detection. Finally, CIoU loss replaced DIoU quicken convergence The experimental results show method 95.4%, floating-point a second (Flops) parameters are 7.4G 5.46M, respectively. Compared traditional YOLOv8, precision algorithm increased 7.7%, recall 1.0%, computational reduced 16.8%, 7.9%. YOLOV8-VCAD obviously better than YOLOv8. paper can quickly accurately detect period. Besides, it reduce cost difficulty vision module equipment, provide technical support automatic intelligence

Language: Английский

Lightweight Salix Cheilophila Recognition Method Based on Improved YOLOv8n DOI
Haotian Ma, Zhigang Liu,

C. C. Pei

et al.

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 11, 2024

Abstract Stumping is an important measure for the care and management of salix cheilophila during its growth. Rapid accurate detection in stumping period desert basis intelligent equipment. However, complex model needs high computing power hardware. It limits deployment application recognition Therefore, this study took areas Shierliancheng, Inner Mongolia Autonomous Region as research object, proposed improved YOLOv8 rapid identification method, named YOLOV8-VCAD. First, lightweight network VanillaNet was used to replace backbone lessen load complexity model. Coordinate attention mechanism embedded extract features by setting location information, which strengthened regression positioning abilities Second, introducing adaptive feature fusion pyramid significantly strengthens model's ability characterize integrate features, improving accuracy performance target detection. Finally, CIoU loss replaced DIoU quicken convergence The experimental results show method 95.4%, floating-point a second (Flops) parameters are 7.4G 5.46M, respectively. Compared traditional YOLOv8, precision algorithm increased 7.7%, recall 1.0%, computational reduced 16.8%, 7.9%. YOLOV8-VCAD obviously better than YOLOv8. paper can quickly accurately detect period. Besides, it reduce cost difficulty vision module equipment, provide technical support automatic intelligence

Language: Английский

Citations

0