MSPB-YOLO: High-Precision Detection Algorithm of Multi-Site Pepper Blight Disease Based on Improved YOLOv8 DOI Creative Commons
Xiaodong Zheng, Zichun Shao, Yile Chen

et al.

Agronomy, Journal Year: 2025, Volume and Issue: 15(4), P. 839 - 839

Published: March 28, 2025

In response to the challenges of low accuracy in traditional pepper blight identification under natural complex conditions, particularly detecting subtle infections on early-stage leaves, stems, and fruits. This study proposes a multi-site disease image recognition algorithm based YOLOv8, named MSPB-YOLO. effectively locates different infection sites peppers. By incorporating RVB-EMA module into model, we can significantly reduce interference from shallow noise high-resolution depth layers. Additionally, introduction RepGFPN network structure enhances model’s capability for multi-scale feature fusion, resulting marked improvement multi-target detection accuracy. Furthermore, optimized CIOU DIOU by integrating center distance bounding boxes loss function; as result, model achieved an impressive [email protected] score 96.4%. represents enhancement 2.2% over original algorithm’s [email protected]. Overall, this provides effective technical support promoting intelligent management prevention strategies

Language: Английский

MSPB-YOLO: High-Precision Detection Algorithm of Multi-Site Pepper Blight Disease Based on Improved YOLOv8 DOI Creative Commons
Xiaodong Zheng, Zichun Shao, Yile Chen

et al.

Agronomy, Journal Year: 2025, Volume and Issue: 15(4), P. 839 - 839

Published: March 28, 2025

In response to the challenges of low accuracy in traditional pepper blight identification under natural complex conditions, particularly detecting subtle infections on early-stage leaves, stems, and fruits. This study proposes a multi-site disease image recognition algorithm based YOLOv8, named MSPB-YOLO. effectively locates different infection sites peppers. By incorporating RVB-EMA module into model, we can significantly reduce interference from shallow noise high-resolution depth layers. Additionally, introduction RepGFPN network structure enhances model’s capability for multi-scale feature fusion, resulting marked improvement multi-target detection accuracy. Furthermore, optimized CIOU DIOU by integrating center distance bounding boxes loss function; as result, model achieved an impressive [email protected] score 96.4%. represents enhancement 2.2% over original algorithm’s [email protected]. Overall, this provides effective technical support promoting intelligent management prevention strategies

Language: Английский

Citations

0