增强的YOLOv8具有轻量级和高效的检测头用于检测大米叶病
Bo Gan1,2, Guolin Pu3, Weiyin Xing4
1Dazhou Vocational and Technical College, Dazhou, 635000, China. ganbcdut@163.com.
Scientific reports
|July 2, 2025
概括
这项研究介绍了G-YOLO,这是一种有效的深度学习模型,用于检测大米叶病. G-YOLO提高了准确性和速度,非常适合在各种设备上实时监测疾病.
科学领域:
- 农业科学 农业科学
- 计算机视觉 计算机视觉
- 机器学习 机器学习
背景情况:
- 准确检测叶病对农业稳定至关重要.
- 像YOLO这样的现有物体检测模型在复杂的环境和计算需求中面临局限性.
- 挑战包括疾病多样性,分布不均和复杂的现场条件.
研究的目的:
- 开发一个优化的物体检测模型,用于精确的,多尺度的叶疾病检测.
- 在特征提取和计算效率方面解决当前YOLO算法的局限性.
- 加强在资源有限的农业设备上部署疾病检测系统.
主要方法:
- 介绍了G-YOLO,这是一个新的架构,集成了轻量级和高效的检测头 (LEDH) 和多尺度空间金字塔聚合快速 (MSPPF).
- LEDH简化了网络结构,以减少计算负载,同时保持准确性.
- MSPPF融合了多层次的特征地图,以改善跨尺度捕获疾病细节.
主要成果:
- 与YOLOv8n.n.相比,G-YOLO在病数据集上表现出优异的性能.
- 实现了4.4%更高的mAP@0.5和3.9%更高的mAP@0.75.
- 每秒 (FPS) 增加了13.1%,这表明处理速度有所提高.
结论:
- G-YOLO在自动检测大米叶疾病方面取得了重大进展.
- 该模型的效率和准确性使其适用于资源有限的环境.
- 这项研究有助于通过先进的人工智能改善作物健康管理.
相关概念视频
Light Acquisition
8.6K
In order to produce glucose, plants need to capture sufficient light energy. Many modern plants have evolved leaves specialized for light acquisition. Leaves can be only millimeters in width or tens of meters wide, depending on the environment. Due to competition for sunlight, evolution has driven the evolution of increasingly larger leaves and taller plants, to avoid shading by their neighbors with contaminant elaboration of root architecture and mechanisms to transport water and nutrients.
8.6K
Plant Breeding and Biotechnology
19.8K
Crop cultivation has a long history in human civilization, with records showing the cultivation of cereal plants beginning at around 8000 BC. This early plant breeding was developed primarily to provide a steady supply of food.
19.8K


