M3VSNET: Unsupervised Multi-Metric Multi-View Stereo Network
(2021) 2021 IEEE International Conference on Image Processing (ICIP) p.3163-3167- Abstract
- The present Multi-view stereo (MVS) methods with supervised learning-based networks have an impressive performance comparing with traditional MVS methods. However, the ground-truth depth maps for training are hard to be obtained and are within limited kinds of scenarios. In this paper, we propose a novel unsupervised multi-metric MVS network, named M 3 VSNet, for dense point cloud reconstruction without any supervision. To improve the robustness and completeness of point cloud reconstruction, we propose a novel multi-metric loss function that combines pixel-wise and feature-wise loss function to learn the inherent constraints from different perspectives of matching correspondences. Besides, we also incorporate the normal-depth consistency... (More)
- The present Multi-view stereo (MVS) methods with supervised learning-based networks have an impressive performance comparing with traditional MVS methods. However, the ground-truth depth maps for training are hard to be obtained and are within limited kinds of scenarios. In this paper, we propose a novel unsupervised multi-metric MVS network, named M 3 VSNet, for dense point cloud reconstruction without any supervision. To improve the robustness and completeness of point cloud reconstruction, we propose a novel multi-metric loss function that combines pixel-wise and feature-wise loss function to learn the inherent constraints from different perspectives of matching correspondences. Besides, we also incorporate the normal-depth consistency in the 3D point cloud format to improve the accuracy and continuity of the estimated depth maps. Experimental results show that M 3 VSNet establishes the state-of-the-arts unsupervised method and achieves better performance than previous supervised MVSNet on the DTU dataset and demonstrates the powerful generalization ability on the Tanks & Temples benchmark with effective improvement. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/36cd0cd7-8941-4dfc-a68b-9b5db8204d91
- author
- Huang, Baichuan LU ; Yi, Hongwei ; Huang, Can ; He, Yijia ; Liu, Jingbin and Liu, Xiao LU
- publishing date
- 2021-09-19
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- host publication
- Proceeding 2021 IEEE International Conference on Image Processing (ICIP)
- pages
- 3163 - 3167
- publisher
- IEEE - Institute of Electrical and Electronics Engineers Inc.
- conference name
- 2021 IEEE International Conference on Image Processing (ICIP)
- conference dates
- 2021-09-19 - 2021-09-22
- external identifiers
-
- scopus:85121115601
- ISBN
- 978-1-6654-4115-5
- DOI
- 10.1109/ICIP42928.2021.9506469
- language
- English
- LU publication?
- no
- id
- 36cd0cd7-8941-4dfc-a68b-9b5db8204d91
- date added to LUP
- 2023-02-21 10:11:02
- date last changed
- 2023-04-11 11:19:53
@inproceedings{36cd0cd7-8941-4dfc-a68b-9b5db8204d91, abstract = {{The present Multi-view stereo (MVS) methods with supervised learning-based networks have an impressive performance comparing with traditional MVS methods. However, the ground-truth depth maps for training are hard to be obtained and are within limited kinds of scenarios. In this paper, we propose a novel unsupervised multi-metric MVS network, named M 3 VSNet, for dense point cloud reconstruction without any supervision. To improve the robustness and completeness of point cloud reconstruction, we propose a novel multi-metric loss function that combines pixel-wise and feature-wise loss function to learn the inherent constraints from different perspectives of matching correspondences. Besides, we also incorporate the normal-depth consistency in the 3D point cloud format to improve the accuracy and continuity of the estimated depth maps. Experimental results show that M 3 VSNet establishes the state-of-the-arts unsupervised method and achieves better performance than previous supervised MVSNet on the DTU dataset and demonstrates the powerful generalization ability on the Tanks & Temples benchmark with effective improvement.}}, author = {{Huang, Baichuan and Yi, Hongwei and Huang, Can and He, Yijia and Liu, Jingbin and Liu, Xiao}}, booktitle = {{Proceeding 2021 IEEE International Conference on Image Processing (ICIP)}}, isbn = {{978-1-6654-4115-5}}, language = {{eng}}, month = {{09}}, pages = {{3163--3167}}, publisher = {{IEEE - Institute of Electrical and Electronics Engineers Inc.}}, title = {{M3VSNET: Unsupervised Multi-Metric Multi-View Stereo Network}}, url = {{http://dx.doi.org/10.1109/ICIP42928.2021.9506469}}, doi = {{10.1109/ICIP42928.2021.9506469}}, year = {{2021}}, }