Journal of Frontiers of Computer Science and Technology ›› 2020, Vol. 14 ›› Issue (11): 1920-1929.DOI: 10.3778/j.issn.1673-9418.1910052

Previous Articles     Next Articles

Deep Optical Flow Learning Networks Combined with Attention Mechanism

ZHOU Haiyun, XIANG Xuezhi, ZHAI Mingliang, ZHANG Rongfang, WANG Shuai   

  1. 1. Institute of Public Security, Nanjing Forest Police College, Nanjing 210023, China
    2. School of Information and Communication Engineering, Harbin Engineering University, Harbin 150001, China
  • Online:2020-11-01 Published:2020-11-09

结合注意力机制的深度学习光流网络

周海赟项学智翟明亮张荣芳王帅   

  1. 1. 南京森林警察学院 治安学院,南京 210023
    2. 哈尔滨工程大学 信息与通信工程学院,哈尔滨 150001

Abstract:

In order to improve the accuracy of deep learning optical flow estimation based on encoder-decoder U-Net, a modified supervised deep optical flow learning network combined with attention mechanism is proposed, which consists of a contracting part and an expanding part. In contracting part, high-level feature information is ex-tracted using a series of convolutional layers, and spatial feature maps are then restored to full resolution by conducting successive deconvolution in expanding part. In this paper, attention mechanism is embedded in U-Net to learn inter-dependencies among the channels so that the channel-wise features can be weighted adaptively, which can enhance the performance of feature extraction. Meanwhile, the proposed network also combines dilated convolution to enlarge the receptive field without changing the size of convolutional kernel. Further, constancy constraints and smoothness constraints from variational method are also adopted so that priori knowledge can be used to improve the accuracy of optical flow estimation. Extensive experiments are conducted on synthesis image sequence datasets and the experi-mental results show the proposed network is effective for improving accuracy of deep learning optical flow estimation.

Key words: optical flow estimation, deep learning, attention mechanism, dilated convolution, prior constraints

摘要:

为提升基于编解码架构的U型网络在深度学习光流估计中的精度,提出了一种结合注意力机制的改进有监督深度学习光流网络。网络由收缩和扩张两部分组成,收缩部分利用一系列卷积层来提取图像之间的高级特征,扩张部分通过反卷积操作将特征图恢复至原始图像分辨率,将通道注意力机制引入U型网络架构中以学习通道之间的相互依赖性,自适应地调整各通道的特征权重,增强网络的特征提取能力。同时,改进的网络还使用了空洞卷积以在卷积核尺寸不变的情况下增大感受野,使用变分光流方法中的恒常约束与平滑约束以进一步利用运动先验知识提升估计效果。最后基于合成图像序列数据集进行了实验验证,实验结果表明所设计的网络能够有效提升深度学习光流估计的准确率。

关键词: 光流估计, 深度学习, 注意力机制, 空洞卷积, 先验约束