引入注意力机制的整体嵌套边缘检测网络

HED Network with Attention Mechanism

边缘检测任务作为计算机视觉的基础性问题,随着深度学习技术的发展取得了显著的进步。回顾与总结了几种经典的手工设计的传统型边缘检测算子,并深入介绍了应用于边缘检测任务的HED卷积神经网络。分析了HED网络的网络结构优缺点,提出了一种引入注意力机制的HED-CBAM网络,经过实验验证有效提升了边缘检测任务的性能。

As a basic problem of computer vision,edge detection has made significant progress with the development of deep learning technology.This paper reviews and summarizes several typical traditional edge detection operators,and deeply introduces the HED network applied to edge detection tasks.After analyzing the advantages and disadvantages of the network structure of HED,a HED-CBAM network is proposed with attention mechanism,which effectively improves the performance of edge detection tasks through experimental verification.