Imperial College London
Portrait Picture

DrStamatiaGiannarou

Faculty of Medicine, Department of Surgery & Cancer

Senior Lecturer
 
 
 
//

Contact

 

+44 (0)20 7594 3492stamatia.giannarou Website

 
 
//

Location

 

413Bessemer BuildingSouth Kensington Campus

//

Summary

 

Publications

Citation

BibTex format

@inproceedings{Huang:2022:10.1109/CVPRW56347.2022.00492,
author = {Huang, B and Zheng, J-Q and Giannarou, S and Elson, DS},
doi = {10.1109/CVPRW56347.2022.00492},
pages = {4459--4466},
publisher = {IEEE},
title = {H-Net: unsupervised attention-based stereo depth estimation leveraging epipolar geometry},
url = {http://dx.doi.org/10.1109/CVPRW56347.2022.00492},
year = {2022}
}
Download

RIS format (EndNote, RefMan)

TY  - CPAPER
AB  - Depth estimation from a stereo image pair has become one of the most explored applications in computer vision, with most previous methods relying on fully supervised learning settings. However, due to the difficulty in acquiring accurate and scalable ground truth data, the training of fully supervised methods is challenging. As an alternative, self-supervised methods are becoming more popular to mitigate this challenge. In this paper, we introduce the H-Net, a deep-learning framework for unsupervised stereo depth estimation that leverages epipolar geometry to refine stereo matching. For the first time, a Siamese autoencoder architecture is used for depth estimation which allows mutual information between rectified stereo images to be extracted. To enforce the epipolar constraint, the mutual epipolar attention mechanism has been designed which gives more emphasis to correspondences of features that lie on the same epipolar line while learning mutual information between the input stereo pair. Stereo correspondences are further enhanced by incorporating semantic information to the proposed attention mechanism. More specifically, the optimal transport algorithm is used to suppress attention and eliminate outliers in areas not visible in both cameras. Extensive experiments on KITTI2015 and Cityscapes show that the proposed modules are able to improve the performance of the unsupervised stereo depth estimation methods while closing the gap with the fully supervised approaches.
AU  - Huang,B
AU  - Zheng,J-Q
AU  - Giannarou,S
AU  - Elson,DS
DO  - 10.1109/CVPRW56347.2022.00492
EP  - 4466
PB  - IEEE
PY  - 2022///
SP  - 4459
TI  - H-Net: unsupervised attention-based stereo depth estimation leveraging epipolar geometry
UR  - http://dx.doi.org/10.1109/CVPRW56347.2022.00492
UR  - https://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=PARTNER_APP&SrcAuth=LinksAMR&KeyUT=WOS:000861612704057&DestLinkType=FullRecord&DestApp=ALL_WOS&UsrCustomerID=a2bf6146997ec60c407a63945d4e92bb
UR  - https://ieeexplore.ieee.org/document/9856940
UR  - http://hdl.handle.net/10044/1/102069
ER  -
Download