RGB image semantic segmentation method and device based on edge metric relation

• Main Authors: ZHANG JIAN, CHEN YEHENG, LI YUEHUA, ZHU SHIQIANG
• Format: Patent
• Language: Chinese; English
• Published: 16.09.2022
• Subjects: CALCULATING; COMPUTING; COUNTING; PHYSICS

The invention discloses an RGB image semantic segmentation method and device based on an edge metric relationship. The method comprises the following steps: constructing a network main body of a semantic segmentation network; extracting separable intermediate layer features from the network main body, and constructing an edge measurement relation module; combining the network main body with the edge measurement relation module, and constructing a semantic segmentation network based on the edge measurement relation; training the semantic segmentation network based on the edge metric relationship; and inputting a to-be-detected RGB image to the trained semantic segmentation network to obtain each semantic category classification score of the to-be-detected RGB image and a semantic segmentation image generated according to the category corresponding to the highest score. 本发明公开了一种基于边缘度量关系的RGB图像语义分割方法及装置，该方法包括：构建语义分割网络的网络主体；从所述网络主体中提取可分离的中间层特征，构建边缘度量关系模块；组合所述网络主体与所述边缘度量关系模块，构建基于边缘度量关系的语义分割网络；对所述基于边缘度量关系的语义分割网络进行训练