CoChat
Sign Up
Cited by 970Robotics and Sensor-Based LocalizationAdvanced Vision and Imaging3D Surveying and Cultural Heritage

Deepvo: towards end-to-end visual odometry with deep recurrent convolutional neural networks

Open access
Wang, Sen, Clark, Ronald, Wen, Hongkai, Trigoni, Niki·2017·2017 IEEE International Conference on Robotics and Automation (ICRA)
DOI 10.1109/icra.2017.7989236
Open this paper in CoChat — free accountOpen PDF

Abstract

This paper studies monocular visual odometry (VO) problem. Most of existing VO algorithms are developed under a standard pipeline including feature extraction, feature matching, motion estimation, local optimisation, etc. Although some of them have demonstrated superior performance, they usually need to be carefully designed and specifically fine-tuned to work well in different environments. Some prior knowledge is also required to recover an absolute scale for monocular VO. This paper presents a novel end-to-end framework for monocular VO by using deep Recurrent Convolutional Neural Networks (RCNNs). Since it is trained and deployed in an end-to-end manner, it infers poses directly from a sequence of raw RGB images (videos) without adopting any module in the conventional VO pipeline. Based on the RCNNs, it not only automatically learns effective feature representation for the VO problem through Convolutional Neural Networks, but also implicitly models sequential dynamics and relations using deep Recurrent Neural Networks. Extensive experiments on the KITTI VO dataset show competitive performance to state-of-the-art methods, verifying that the end-to-end Deep Learning technique can be a viable complement to the traditional VO systems.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1109/icra.2017.7989236

Cites

29 of 29 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Deep-learning-based visual odometry accuracy is inferior to traditional methods because it ignores geometric information.

    “More surprisingly, they show that the welltrained networks enable scale-consistent predictions over long videos, while the accuracy is still inferior to traditional methods because of ignoring geometric information.”

    From DF-VO: What Should Be Learnt for Visual Odometry? (2021) · cites this paper as Wang et al. 2017

    In B. VO Results (p. 6), DeepVO performs worse than traditional stereo VISO2 but generally better than traditional monocular VISO2, so the claim overgeneralizes the accuracy comparison. B. Methods based on Learning (p. 2) says learning-based methods do not explicitly apply geometric theory, but the paper does not say that ignoring geometry causes the accuracy gap; it suggests insufficient high-speed training data for some errors. (LLM, conf=0.87)

    Experimental — accuracy not yet measured

  • End-to-end deep learning approaches can estimate the 6 degrees of freedom (DoF) poses directly from a sequence of RGB images without engineering effort to design the modules in the traditional VO pipeline.

    “On the opposing side, end-to-end approaches can estimate the 6 degrees of freedom (DoF) poses directly from a sequence of RGB images, that is, it does not need engineering effort to design the modules in the traditional VO pipeline [4].”

    From Transformer-Based Model for Monocular Visual Odometry: A Video Understanding Approach (2023) · cites this paper as [4]

  • Traditional geometry-based monocular visual odometry methods require considerable engineering effort to be fine-tuned for each specific application.

    “This problem is called monocular visual odometry and often relies on geometric approaches that require considerable engineering effort for a specific scenario.”

    From Transformer-Based Model for Monocular Visual Odometry: A Video Understanding Approach (2023) · cites this paper as [4]

  • The ODN crater detections are robust to disturbances or modifications of image quality due to Gaussian noise, shot noise, brightness levels, and different illumination conditions.

    “the amount of crater detections that ODN-detector produces is robust to disturbances or modifications of the image quality due to Gaussian noise, shot noise, brightness levels, as well as different illumination conditions.”

    From Optical navigation for Lunar landing based on Convolutional Neural Network crater detector (2022) · cites this paper as [16]

    The Training and Testing section (p. 5) describes DeepVO visual odometry, not ODN crater detection. The VO Results (p. 7) suggest augmenting future training data with noise, but do not report tests of Gaussian noise, shot noise, brightness, or illumination on crater detections. Searches for ODN crater detections and those image-quality conditions found no supporting passage in our copy. (LLM, conf=0.78)

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

Flagged first, then by citations

Research works printed but not linked: 2

  1. [15][15] R. Roberts, H. Nguyen, N. Krishnamurthi, and T. Balch, 'Memorybased learning for visual odometry,' in Proceedings of IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2008, pp. 47-52. -
  2. [26][26] W. Zaremba and I. Sutskever, 'Learning to execute,' arXiv preprint arXiv:1410.4615 , 2014. Fig. 8. Trajectories of VO results on the testing Sequence 11, 12, 15, 17, 18 and 19 of the KITTI VO benchmark (no ground truth is available for these testing sequences). The DeepVO model used is trained on the whole training dataset of the KITTI VO benchmark. <!-- image --> · arXiv

Cited by

Sorted by citations

Related

Authors
Wang, S., Clark, R., Wen, H., & Trigoni, N.
Affiliations
University of Oxford (GB)
ORCID
Venue
2017 IEEE International Conference on Robotics and Automation (ICRA)
Publisher
IEEE
Published
2017-05
Pages
2043-2050
ISSN
1050-4729 (ISSN-L)
DOI
10.1109/icra.2017.7989236
OpenAlex
W2598706937
Cited by
970
References
29
Citations by year
2026: 63 · 2025: 120 · 2024: 104 · 2023: 123 · 2022: 130 · 2021: 131 · 2020: 125 · 2019: 116 · 2018: 52 · 2017: 6
FWCI
143.15
Type
conference paper
Language
en
Open access
Locations
  • Proceedings - IEEE International Conference on Robotics and Automation/Proceedings (publishedVersion)
  • arXiv (Cornell University) (publishedVersion)
  • Warwick Research Archive Portal (University of Warwick) (submittedVersion)
  • arXiv (Cornell University)
Funders
Engineering and Physical Sciences Research Council (EP/M019918/1)
Topics
  • Physical Sciences › Engineering › Aerospace Engineering › Robotics and Sensor-Based Localization
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Vision and Imaging
  • Physical Sciences › Earth and Planetary Sciences › Geology › 3D Surveying and Cultural Heritage
Keywords
monocular visual odometry, end-to-end learning, recurrent convolutional neural network, pose estimation, feature representation learning

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review