计算机与数字工程2023，Vol.51Issue(12)：2966-2970,5.DOI:10.3969/j.issn.1672-9722.2023.12.036
基于多分支卷积的单目深度估计方法研究
Research on Monocular Depth Estimation Based on Diverse Branch Block
摘要
Abstract
Monocular depth estimation currently has challenges such as poor accuracy and blurred object boundary depth pre-diction.In response to the above problems,this paper proposes a monocular depth estimation algorithm based on diverse-branch convolution,which uses complex convolution structure to extract richer scenes in the scene semantic information.The model uses four-branch convolution to replace the original single-branch convolution in the training phase,and the weight parameters of the di-verse-branch convolution can be transplanted to the original single-branch network during test deployment,so that no additional in-ference time is added to the network model during the testing and use phases.In the test comparison of public datasets,the depth map results predicted by the method proposed in this paper are clearer,and can effectively deal with areas such as object boundar-ies in the picture.The experimental results show that the method proposed in this paper has certain effectiveness.关键词
卷积神经网络/自监督学习/单目深度估计/多分支卷积Key words
convolutional neural network/self-supervised learning/monocular depth estimation/diverse-branch convolu-tion分类
信息技术与安全科学引用本文复制引用
印雅萌,周嘉麒,王指辉..基于多分支卷积的单目深度估计方法研究[J].计算机与数字工程,2023,51(12):2966-2970,5.
