The DOI matches its Crossref record; 125 of the 127 references it cites were found in a registry or an index.
Approaches, Challenges, and Applications for Deep Visual Odometry: Toward Complicated and Emerging Areas
Open accessAbstract
Visual odometry (VO) is a prevalent way to deal with the relative localization problem, which is becoming increasingly mature and accurate, but it tends to be fragile under challenging environments. Comparing with classical geometry-based methods, deep-learning-based methods can automatically learn effective and robust representations, such as depth, optical flow, feature, ego-motion, etc., from data without explicit computation. Nevertheless, there still lacks a thorough review of the recent advances of deep-learning-based VO (Deep VO). Therefore, this article aims to gain a deep insight on how deep learning can profit and optimize the VO systems. We first screen out a number of qualifications, including accuracy, efficiency, scalability, dynamicity, practicability, and extensibility, and employ them as the criteria. Then, using the offered criteria as the uniform measurements, we detailedly evaluate and discuss how deep learning improves the performance of VO from the aspects of depth estimation, feature extraction and matching, and pose estimation. We also summarize the complicated and emerging areas of Deep VO, such as mobile robots, medical robots, augmented and virtual reality, etc. Through the literature decomposition, analysis, and comparison, we finally put forward a number of open issues and raise some future research directions in this field.
Overview
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.
Integrity signals
Registry
Cites
Statistics
Verification
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.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Are we ready for autonomous driving? The KITTI vision benchmark suite
Geiger, A., Lenz, P., & Urtasun, R. (2012) · 2012 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2012.6248074 · Cited by 14,897 · cited 1×
“For the former, we use RMSE (Root Mean Square Error) [23] to measure the bias of ground truth and estimation, which directly describes algorithm accuracy.”
MethodORB-SLAM: A Versatile and Accurate Monocular SLAM System
Mur-Artal, R., Montiel, J. M. M., & Tardos, J. D. (2015) · IEEE Transactions on Robotics · doi:10.1109/tro.2015.2463671 · Cited by 6,778 · cited 2×
“use root-mean-square error (RMSE, defined in [28]) or average trajectory error (ATE, defined in [4]) to measure the bias of ground truth and estimation, which directly describes algorithm accuracy.”
BackgroundMethodLSD-SLAM: Large-scale direct monocular SLAM
Engel, J., Schöps, T., & Cremers, D. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-10605-2_54 · Cited by 3,851
Past, Present, and Future of Simultaneous Localization and Mapping: Toward the Robust-Perception Age
Cadena, C., Carlone, L., Carrillo, H., Latif, Y., Scaramuzza, D., Neira, J., Reid, I., & Leonard, J. J. (2016) · IEEE Transactions on Robotics · doi:10.1109/tro.2016.2624754 · Cited by 3,596 · cited 1×
“Most recent review papers focus on geometrybased methods, less on deep-learning-based approaches [16], [18], [19].”
BackgroundDepth Map Prediction from a Single Image using a Multi-Scale Deep Network
Eigen, D., Puhrsch, C., & Fergus, R. (2014) · arXiv · doi:10.48550/arxiv.1406.2283 · Cited by 3,442 · cited 1×
“1) Supervised Learning Methods: For supervised learning, the use of supervised-based methods was first proposed in [29].”
MethodUnsupervised Monocular Depth Estimation with Left-Right Consistency
Godard, C., Aodha, O. M., & Brostow, G. J. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.699 · Cited by 3,270 · cited 2×
“[46] introduced a method that can simultaneously infer disparities that warp the left images to match the right one or warp the right image to match the left one, thereby obtaining better depth estimation by enforcing left–right depth consistency.”
BackgroundMethodEngel, J. J., Koltun, V., & Cremers, D. (2017) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2017.2658577 · Cited by 2,995
Unsupervised Learning of Depth and Ego-Motion from Video
Zhou, T., Brown, M., Snavely, N., & Lowe, D. G. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.700 · Cited by 2,910 · cited 1×
“[47] first synthesized a new image of the scene from different views of the input image.”
BackgroundPoseNet: A Convolutional Network for Real-Time 6-DOF Camera Relocalization
Kendall, A., Grimes, M., & Cipolla, R. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.336 · Cited by 2,489 · cited 1×
“1) Supervised Methods: PoseNet [83] was the work that the first to utilize CNN to regress 6 DOF camera pose from the monocular image in an end-to-end manner without graph optimization.”
MethodUnsupervised CNN for Single View Depth Estimation: Geometry to the Rescue
Garg, R., B.G., V. K., Carneiro, G., & Reid, I. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46484-8_45 · Cited by 1,632 · cited 1×
“The method using unsupervised learning was first introduced in [45], which estimates the depth information from single-view images.”
Method
Flagged first, then by citations
2 of the 127 references couldn’t be matched to a known paper.
Research works printed but not linked: 13
- Campos, 2020, ORB-SLAM3: An accurate open-source library for visual, visual-inertial and multi-map SLAM
- 10.1007/978-3-030-01237-3_40 · DOI
- Ma, 2020, Progress in research on monocular visual odometry of autonomous vehicles, J. Jilin Univ., 50, 765
- 10.1145/3177853 · DOI
- Zhang, 2019, DFineNet: Ego-motion estimation and depth refinement from sparse, noisy depth input with RGB guidance
- Ramirez, 2018, Geometry meets semantics for semi-supervised monocular depth estimation
- Tosi, 2019, Learning monocular depth estimation infusing traditional stereo knowledge, 10.1109/CVPR.2019.01003 · DOI
- Mishchuk, 2017, Working hard to know your neighbor’s margins: Local descriptor learning loss, 30
- Cieslewski, 2018, Matching features without descriptors: Implicitly matched interest points
- Jiao, 2018, MagicVO: End-to-end monocular visual odometry through deep bi-directional recurrent convolutional neural network
- 10.1007/978-3-030-29911-8_35 · DOI
- Vijayanarasimhan, 2017, SfM-Net: Learning of structure and motion from video
- Wei, 2019, Instant motion tracking and its applications to augmented reality
Cited by
Sorted by citations
Items 101–114 of 114
Object-level Visual Semantic SLAM Based on Geometric Constraints and Outlier Removal
Wang, L., Feng, W., & Li, H. (2025) · Chinese Control and Decision Conference (CCDC) · doi:10.1109/ccdc65474.2025.11090992
“Semantic visual SLAM has gained attention to address these challenges via integrating semantic information for enhanced map construction and localization [3], [4].”
Visual Slam in the Heterogeneous Intelligence Era: A SurveyPreprint
Su, S., He, X., Meng, X., Zhang, Y., Yin, W., Mo, L., & Yu, F. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5383002
A Multi-Level Feature Fusion-Based Network for Motion Estimation
Wang, H., Gao, S., Yu, S., Sun, R., & Sun, L. (2025) · IEEE International Conference on Cyborg and Bionic Systems (CBS) · doi:10.1109/cbs65871.2025.11267742
A Spiking Transfer Learning Method for Event Camera Based on Structural Decoupling
Kan, P., Zhan, Q., Tao, R., Li, Z., Qin, Z., & Liu, G. (2025) · International Conference on Machine Intelligence and Nature-Inspired Computing (MIND) · doi:10.1109/mind67540.2025.11351579
“Some works have explored zero-shot learning to increase the utilization of limited event data [10], or directly used simulators [11] to convert frame-based data into event stream data for downstream tasks [12].”
Choi, J. H., Kang, C. H., & Kim, S. Y. (2026) · International Journal of Control Automation and Systems · doi:10.1007/s12555-026-00029-2
Geometry Knowledge-Embedded Self-supervised Deep Monocular Visual Odometry for Autonomus DrivingPreprint
Tian, W., Zheng, D., Zhao, Y., Zhang, F., Gui, G., & Gui, W. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6318426
Yuan, Z., Shen, Y., Wang, J., & Wang, Y. (2026) · Research Square · doi:10.21203/rs.3.rs-9365602/v1
Robust and low-complexity feature matching for UAV SLAM in repetitive environments
Quach, C. H., Perry, S., Lê, H. V., Pham, T., Tran, T. H., & Phung, M. D. (2026) · Results in Engineering · doi:10.1016/j.rineng.2026.110864
Geometry knowledge-embedded self-supervised deep monocular visual odometry for autonomous driving
Zheng, D., Tian, W., Zhao, Y., Zhang, F., Gui, G., & Gui, W. (2026) · Knowledge-Based Systems · doi:10.1016/j.knosys.2026.116256
Tong, B., Chen, W., Du, L., Wang, H., Yan, X., & Si, F. (2026) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2026.3711340
Adaptive Hybrid Patch Sampling for Deep Patch Visual Odometry
Lv, G., Li, Z., Shu, S., Zongyuan, L., Wang, W., Zhang, R., & Zhao, L. (2026) · International Conference on Artificial Intelligence, Big Data and Algorithms (CAIBDA) · doi:10.1109/caibda70336.2026.11621272
A RGB-D SLAM method based on contact experience in dynamic environment
Zhang, J., Huang, K., Geng, H., Song, C., & Zhang, M. (2026) · Autonomous Robots · doi:10.1007/s10514-026-10268-1
“With the maturation of object detection and semantic segmentation technologies, many studies have applied them to vSLAM systems (Ran et al., 2023; Wang et al., 2022).”
Transformer-Based End-to-End Model for Monocular Visual Odometry on Urban Traffic Scenes
Pontarolo, G., & Todt, E. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-39898-7_33
Transformer-Based Model for Monocular Visual Odometry: A Video Understanding Approach
Françani, A. O., & Maximo, M. R. O. A. (2023) · arXiv · doi:10.48550/arxiv.2305.06121 · Cited by 38
“As we observed in our literature review, there are only a few supervised end-to-end models for 6-DoF pose estimation [27] in the context of visual odometry with imagery data, since most researched approaches are hybrid that still use geometric modeling in their VO components, or they deal with VIO,…”
Background
Metadata
- Authors
- Wang, K., Ma, S., Chen, J., Ren, F., & Lu, J.
- Affiliations
- Chongqing University (CN)
- Chongqing Normal University (CN)
- Changan Auto Company, Intelligent Vehicle R&D Institute, Chongqing, China
- Ford Motor Company (United States) (US)
- Venue
- IEEE Transactions on Cognitive and Developmental Systems
- Published
- 2022-03
- Volume
- 14
- Issue
- 1
- Pages
- 35-49
- ISSN
- 2379-8920 (ISSN-L)
- 2379-8939
- OpenAlex
- W3083208433
- Cited by
- 114
- References
- 127
- Citations by year
- 2026: 10 · 2025: 15 · 2024: 35 · 2023: 25 · 2022: 19
- FWCI
- 8.11
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Transactions on Cognitive and Developmental Systems (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- National Natural Science Foundation of China (51605054)
- National Key Research and Development Program of China (SQ2020YFF0410766)
- Natural Science Foundation of Chongqing (cstc2020jcyj-msxmX0575)
- Chongqing Technology Innovation and Application Development Project (cstc2020jscxmsxmX0109)
- State Key Laboratory of Vehicle NVH and Safety Technology (NVHSKL-202008, NVHSKL-202010)
- Fundamental Research Funds for the Central Universities (2020CDJ-LHZZ-042)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › 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
- visual odometry, deep learning, relative localization, depth estimation, feature matching, pose estimation, mobile robots, medical robots, augmented reality, virtual reality
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
