UnDeepVO: Monocular Visual Odometry through Unsupervised Deep Learning

Ruihao Li; Sen Wang; Zhiqiang Long; Dongbing Gu

doi:10.1109/ICRA.2018.8461251

UnDeepVO: Monocular Visual Odometry through Unsupervised Deep Learning

Ruihao Li, Sen Wang, Zhiqiang Long, Dongbing Gu

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

401 Citations (Scopus)

119 Downloads (Pure)

Abstract

We propose a novel monocular visual odometry (VO) system called UnDeepVO in this paper. UnDeepVO is able to estimate the 6-DoF pose of a monocular camera and the depth of its view by using deep neural networks. There are two salient features of the proposed UnDeepVo:one is the unsupervised deep learning scheme, and the other is the absolute scale recovery. Specifically, we train UnDeepVoby using stereo image pairs to recover the scale but test it by using consecutive monocular images. Thus, UnDeepVO is a monocular system. The loss function defined for training the networks is based on spatial and temporal dense information. A system overview is shown in Fig. 1. The experiments on KITTI dataset show our UnDeepVO achieves good performance in terms of pose accuracy.

Original language	English
Title of host publication	2018 IEEE International Conference on Robotics and Automation (ICRA)
Publisher	IEEE
Pages	7286-7291
Number of pages	6
ISBN (Electronic)	9781538630815
DOIs	https://doi.org/10.1109/ICRA.2018.8461251
Publication status	Published - 13 Sept 2018

Publication series

Name	International Conference on Robotics and Automation
Publisher	IEEE
ISSN (Electronic)	2577-087X

Access to Document

10.1109/ICRA.2018.8461251

Download pdf
© 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
Accepted author manuscript, 1.44 MB

https://arxiv.org/pdf/1709.06841.pdf

Cite this

@inproceedings{3bd2d6a7f9b5424298c2272059cc8e3d,

title = "UnDeepVO: Monocular Visual Odometry through Unsupervised Deep Learning",

abstract = "We propose a novel monocular visual odometry (VO) system called UnDeepVO in this paper. UnDeepVO is able to estimate the 6-DoF pose of a monocular camera and the depth of its view by using deep neural networks. There are two salient features of the proposed UnDeepVo:one is the unsupervised deep learning scheme, and the other is the absolute scale recovery. Specifically, we train UnDeepVoby using stereo image pairs to recover the scale but test it by using consecutive monocular images. Thus, UnDeepVO is a monocular system. The loss function defined for training the networks is based on spatial and temporal dense information. A system overview is shown in Fig. 1. The experiments on KITTI dataset show our UnDeepVO achieves good performance in terms of pose accuracy.",

author = "Ruihao Li and Sen Wang and Zhiqiang Long and Dongbing Gu",

year = "2018",

month = sep,

day = "13",

doi = "10.1109/ICRA.2018.8461251",

language = "English",

series = "International Conference on Robotics and Automation",

publisher = "IEEE",

pages = "7286--7291",

booktitle = "2018 IEEE International Conference on Robotics and Automation (ICRA)",

address = "United States",

}

TY - GEN

T1 - UnDeepVO: Monocular Visual Odometry through Unsupervised Deep Learning

AU - Li, Ruihao

AU - Wang, Sen

AU - Long, Zhiqiang

AU - Gu, Dongbing

PY - 2018/9/13

Y1 - 2018/9/13

N2 - We propose a novel monocular visual odometry (VO) system called UnDeepVO in this paper. UnDeepVO is able to estimate the 6-DoF pose of a monocular camera and the depth of its view by using deep neural networks. There are two salient features of the proposed UnDeepVo:one is the unsupervised deep learning scheme, and the other is the absolute scale recovery. Specifically, we train UnDeepVoby using stereo image pairs to recover the scale but test it by using consecutive monocular images. Thus, UnDeepVO is a monocular system. The loss function defined for training the networks is based on spatial and temporal dense information. A system overview is shown in Fig. 1. The experiments on KITTI dataset show our UnDeepVO achieves good performance in terms of pose accuracy.

AB - We propose a novel monocular visual odometry (VO) system called UnDeepVO in this paper. UnDeepVO is able to estimate the 6-DoF pose of a monocular camera and the depth of its view by using deep neural networks. There are two salient features of the proposed UnDeepVo:one is the unsupervised deep learning scheme, and the other is the absolute scale recovery. Specifically, we train UnDeepVoby using stereo image pairs to recover the scale but test it by using consecutive monocular images. Thus, UnDeepVO is a monocular system. The loss function defined for training the networks is based on spatial and temporal dense information. A system overview is shown in Fig. 1. The experiments on KITTI dataset show our UnDeepVO achieves good performance in terms of pose accuracy.

U2 - 10.1109/ICRA.2018.8461251

DO - 10.1109/ICRA.2018.8461251

M3 - Conference contribution

T3 - International Conference on Robotics and Automation

SP - 7286

EP - 7291

BT - 2018 IEEE International Conference on Robotics and Automation (ICRA)

PB - IEEE

ER -

UnDeepVO: Monocular Visual Odometry through Unsupervised Deep Learning

Abstract

Publication series

Access to Document

Fingerprint

Cite this