Depth Completion using Geometry-Aware Embedding

被引:23
作者
Du, Wenchao [1 ]
Chen, Hu [1 ]
Yang, Hongyu [1 ]
Zhang, Yi [1 ]
机构
[1] Sichuan Univ, Coll Comp Sci, Chengdu 610065, Peoples R China
来源
2022 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION, ICRA 2022 | 2022年
关键词
NETWORK; SPARSE;
D O I
10.1109/ICRA46639.2022.9811556
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Exploiting internal spatial geometric constraints of sparse LiDARs is beneficial to depth completion, however, has been not explored well. This paper proposes an efficient method to learn geometry-aware embedding, which encodes the local and global geometric structure information from 3D points, e.g., scene layout, object's sizes and shapes, to guide dense depth estimation. Specifically, we utilize the dynamic graph representation to model generalized geometric relationship from irregular point clouds in a flexible and efficient manner. Further, we joint this embedding and corresponded RGB appearance information to infer missing depths of the scene with well structure-preserved details. The key to our method is to integrate implicit 3D geometric representation into a 2D learning architecture, which leads to a better trade-off between the performance and efficiency. Extensive experiments demonstrate that the proposed method outperforms previous works and could reconstruct fine depths with crisp boundaries in regions that are over-smoothed by them. The ablation study gives more insights into our method that could achieve significant gains with a simple design, while having better generalization capability and stability. The code is available at https://github.com/Wenchao-Du/GAENet.
引用
收藏
页码:8680 / 8686
页数:7
相关论文
共 44 条
[1]  
[Anonymous], 2020, 2019 16 INT C MACH V, DOI DOI 10.1080/19491247.2019.1682234
[2]   Transformer Tracking [J].
Chen, Xin ;
Yan, Bin ;
Zhu, Jiawen ;
Wang, Dong ;
Yang, Xiaoyun ;
Lu, Huchuan .
2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021, 2021, :8122-8131
[3]  
Cheng X., 2018, ECCV, P103
[4]  
de Silva C. W., 2019, LINKED DYNAMIC GRAPH
[5]   A Novel Loop: Mutual Regulation Between Epigenetic Modification and the Circadian Clock [J].
Du, Shenxiu ;
Chen, Liang ;
Ge, Liangfa ;
Huang, Wei .
FRONTIERS IN PLANT SCIENCE, 2019, 10
[6]   Confidence Propagation through CNNs for Guided Sparse Depth Regression [J].
Eldesokey, Abdelrahman ;
Felsberg, Michael ;
Khan, Fahad Shahbaz .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2020, 42 (10) :2423-2436
[7]  
Hawe S, 2011, IEEE I CONF COMP VIS, P2126, DOI 10.1109/ICCV.2011.6126488
[8]   Visualization of Convolutional Neural Networks for Monocular Depth Estimation [J].
Hu, Junjie ;
Zhang, Yan ;
Okatani, Takayuki .
2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2019), 2019, :3868-3877
[9]  
Hu M., 2021, 2021 IEEE INT C ROB
[10]   HMS-Net: Hierarchical Multi-Scale Sparsity-Invariant Network for Sparse Depth Completion [J].
Huang, Zixuan ;
Fan, Junming ;
Cheng, Shenggan ;
Yi, Shuai ;
Wang, Xiaogang ;
Li, Hongsheng .
IEEE TRANSACTIONS ON IMAGE PROCESSING, 2020, 29 :3429-3441