A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Road disease detection algorithm based on improved YOLOv8
The invention discloses a road disease detection algorithm based on improved YOLOv8, and the algorithm comprises the steps: introducing a DLKA into a C2f module, and specifically comprises the steps: replacing a bottleneck in C2f with a bottleneck DLKA, namely replacing conv in the original bottleneck with a deformconv; according to the method, three MSDA modules are added to a check part, an input feature map is divided into different heads, feature extraction is performed on the heads by using expansion convolution with different expansion rates, finally all feature extraction results are fused, the MSDA mechanism can fully extract sufficient information of objects with large, medium and small sizes, and the accuracy of feature extraction is improved. The calculated amount is not increased too much, and the method can well adapt to detection of objects with various scales, such as pavement diseases.
本发明公开了一种基于改进的YOLOv8的道路病害检测算法，包括将DLKA引入C2f模块中，具体方法是用bottleneck DLKA替换C2f中的bottleneck，即用deformconv替换了原有bottleneck中的conv；本发明在neck部分添加了三处MSDA模块，通过将输入的特征图分为不同的头部，并在它们上面使用不同扩张率的扩张卷积进行特征提取，最后将所有的特征提取结果融合，MSDA机制可以充分提取到大、中、小尺寸的物体的充分信息，又不会过多的增加计算量，可以很好的适应路面病害这种尺度多样的物体的检测。
Road disease detection algorithm based on improved YOLOv8
The invention discloses a road disease detection algorithm based on improved YOLOv8, and the algorithm comprises the steps: introducing a DLKA into a C2f module, and specifically comprises the steps: replacing a bottleneck in C2f with a bottleneck DLKA, namely replacing conv in the original bottleneck with a deformconv; according to the method, three MSDA modules are added to a check part, an input feature map is divided into different heads, feature extraction is performed on the heads by using expansion convolution with different expansion rates, finally all feature extraction results are fused, the MSDA mechanism can fully extract sufficient information of objects with large, medium and small sizes, and the accuracy of feature extraction is improved. The calculated amount is not increased too much, and the method can well adapt to detection of objects with various scales, such as pavement diseases.
本发明公开了一种基于改进的YOLOv8的道路病害检测算法，包括将DLKA引入C2f模块中，具体方法是用bottleneck DLKA替换C2f中的bottleneck，即用deformconv替换了原有bottleneck中的conv；本发明在neck部分添加了三处MSDA模块，通过将输入的特征图分为不同的头部，并在它们上面使用不同扩张率的扩张卷积进行特征提取，最后将所有的特征提取结果融合，MSDA机制可以充分提取到大、中、小尺寸的物体的充分信息，又不会过多的增加计算量，可以很好的适应路面病害这种尺度多样的物体的检测。
Road disease detection algorithm based on improved YOLOv8
一种基于改进的YOLOv8的道路病害检测算法
HAN ZHIBIN (author) / CAI YUTONG (author) / LI YAHANG (author) / LIU ANQI (author) / ZHAO YIRAN (author) / LIN CIYUN (author)
2024-08-16
Patent
Electronic Resource
Chinese
GLU-YOLOv8: An Improved Pest and Disease Target Detection Algorithm Based on YOLOv8
| DOAJ | 2024
Fire Detection Based on Improved YOLOv8 Network Mode
| IEEE | 2024
YOLOv8s-SNC: An Improved Safety-Helmet-Wearing Detection Algorithm Based on YOLOv8
| DOAJ | 2024