|
| 一种用于深度补全的双分支引导网络 |
| A dual-branch guided network for depth completion |
| 投稿时间：2022-12-17 |
| DOI：10.3969/j.issn.1005-5630.2023.005.008 |
| 中文关键词: 深度补全 多数据引导 卷积神经网络 谱残差块 注意力机制 |
| 英文关键词:depth completion multiple data guidance convolution neural network spectral residual block attention mechanism |
| 基金项目:国家自然科学基金重点项目（92048205）；国家留学基金（202008310014） |
|
| 摘要点击次数: 1957 |
| 全文下载次数: 1617 |
| 中文摘要: |
| 深度信息在机器人、自动驾驶等领域中有着重要作用，通过深度传感器获取的深度图较为稀疏，研究人员为了补全缺失的深度信息提出了大量方法。但现有方法大多是针对不透明对象，基于卷积神经网络的强大表征能力，设计了一个双分支引导的编解码结构网络模型，通过针对透明物体的以掩码图为引导的编码分支，提升网络对透明物体特征信息的提取能力，并且使用谱残差块连接编解码部分，提高了网络训练稳定性及获取物体结构信息的能力，除此之外，还加入了注意力机制以提升网络空间和语义信息的特征建模能力。该网络在两个数据集上都达到了领先的效果。 |
| 英文摘要: |
| Depth information plays an important role in the fields of robotics and autonomous driving. The depth map obtained by the depth sensor is relatively sparse. Researchers have proposed a large number of methods to complement the missing depth values. However, most of the existing methods aim at opaque objects. Based on the powerful representation ability of convolution neural network, this paper designed a dual-branch-guided encoder-decoder structure network. Through mask-guided branch for transparent objects, it improves the ability of the network to extract feature information of transparent objects. And spectral residual blocks improves the stability of network in training process and the ability to obtain object structure information. In addition, attention mechanism is added to improve the feature modeling ability of network space and semantic information. The network achieves state-of-the-art results on all two datasets. |
| HTML 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
