The DOI matches its Crossref record; 29 of the 29 references it cites were found in a registry or an index.
DeepVO: Towards end-to-end visual odometry with deep Recurrent Convolutional Neural Networks
Open accessAbstract
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.
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.
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?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“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
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“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]
The cited source says what the claim attributes to it.
In the citing paper
“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 check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“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]
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.
Citations
Cites
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 1×
“There have been some popular and powerful DNN architectures, such as VGGNet [22] and GoogLeNet [23], developed for computer vision tasks, producing remarkable performance.”
MethodGoing deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562
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 2×
“In this section, we evaluate the proposed end-to-end monocular VO approach on the well-known KITTI VO/SLAM benchmark [3].”
BackgroundMethod(2016) · Cited by 11,679
ORB-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 1×
“Examples include keyframe base PTAM [9] and ORB-SLAM [10].”
BackgroundParallel Tracking and Mapping for Small AR Workspaces
Klein, G., & Murray, D. (2007) · 2007 6th IEEE and ACM International Symposium on Mixed and Augmented Reality · doi:10.1109/ismar.2007.4538852 · Cited by 4,346 · cited 1×
“Examples include keyframe base PTAM [9] and ORB-SLAM [10].”
BackgroundFlownet: Learning optical flow with convolutional networks
Dosovitskiy, A., Fischer, P., Ilg, E., Hausser, P., Hazirbas, C., Golkov, V., Smagt, P. V. D., Cremers, D., & Brox, T. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.316 · Cited by 4,313 · cited 2×
“In order to reduce both the training time and data required to converge, the CNN is based on a pre-trained FlowNet model [24].”
MethodMonoSLAM: Real-time single camera SLAM
Davison, A. J., Reid, I. D., Molton, N. D., & Stasse, O. (2007) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2007.1049 · Cited by 3,959 · cited 1×
“To mitigate this problem, visual Simultaneous Localisation and Mapping (SLAM) or Structure from Motion (SfM) can be adopted to maintain a feature map for drift correction along with pose estimation [8].”
MethodORB-SLAM: A Versatile and Accurate Monocular SLAM System
Mur-Artal, R., Montiel, J. M. M., & Tardos, J. D. (2015) · Zaguan (Universidad de Zaragoza) · Cited by 3,153
Multiple view geometry in computer vision
Opower, H. (2002) · Optics and Lasers in Engineering · doi:10.1016/s0143-8166(01)00145-2 · Cited by 3,044
Flagged first, then by citations
Research works printed but not linked: 1
- zaremba, 2014, Learning to execute
Cited by
Sorted by citations
Items 301–400 of 969
Learned Monocular Depth Priors in Visual-Inertial Initialization
Zhou, Y., Kar, A., Turner, E., Kowdle, A., Guo, C. X., DuToit, R. C., & Tsotsos, K. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20047-2_32 · Cited by 15
Pseudo-LiDAR for Visual Odometry
Miao, Y., Deng, H., Jiang, C., Feng, Z., Wu, X., Wang, G., & Wang, H. (2023) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2023.3315416 · Cited by 15
“Methods of visual odometry can be mainly divided into geometry-based methods [8], [9] and deep learning-based methods [2], [10].”
BackgroundMethodInfluentialSelf-supervised deep monocular visual odometry and depth estimation with observation variation
Zhao, W., Wang, Y., Wang, Z., Li, R., Xiao, P., Wang, J., & Guo, R. (2023) · Displays · doi:10.1016/j.displa.2023.102553 · Cited by 15
A Review of Crowdsourcing Update Methods for High-Definition Maps
Guo, Y., Zhou, J. Y., Li, X., Tang, Y., & Lv, Z. (2024) · ISPRS International Journal of Geo-Information · doi:10.3390/ijgi13030104 · Cited by 15
Deep Intersection Classification Using First and Third Person Views
Takeda, K., & Kanji, T. (2019) · IEEE Intelligent Vehicles Symposium · doi:10.1109/ivs.2019.8813859 · Cited by 15
“These are typically recognized by long-short-termmemory (LSTM) [1].”
BackgroundMethodInfluentialSemi-Supervised Monocular Depth Estimation Based on Semantic Supervision
Yue, M., Fu, G., Wu, M., & Wang, H. (2020) · Journal of Intelligent & Robotic Systems · doi:10.1007/s10846-020-01205-0 · Cited by 15
“Wang et al. [12] proposed a visual odometry method based on circular convolutional neural network, which combined the LSTM unit with the convolutional layer to achieve good results.”
MethodBenchmarking Cameras for Open VSLAM Indoors
Chappellet, K., Caron, G., Kanehiro, F., Sakurada, K., & Kheddar, A. (2021) · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · doi:10.1109/icpr48806.2021.9413278 · Cited by 15
“…of localization and mapping thanks to a better handling of partial Stereo Dual [4], [5], [6], [7] [8] [9], [3] [10], [11], [12] ( a ) [13], [14], [15], [16], [17] [18], [19], [20], [21] [22], [23], [24] ( b ) [25] ( c ) [26] ( c ) [2] ( a ) at least for learning, ( b ) as diverging optical axes,…”
POU-SLAM: Scan-to-Model Matching Based on 3D Voxels
Jiang, J., Wang, J., Wang, P., & Chen, Z. (2019) · Applied Sciences · doi:10.3390/app9194147 · Cited by 15
Unsupervised monocular visual odometry with decoupled camera pose estimation
Lin, L., Wang, W., Luo, W., Song, L., & Zhou, W. (2021) · Digital Signal Processing · doi:10.1016/j.dsp.2021.103052 · Cited by 15
A compact CNN approach for drone localisation in autonomous drone racing
Cocoma-Ortega, J. A., & Martínez-Carranza, J. (2021) · Journal of Real-Time Image Processing · doi:10.1007/s11554-021-01162-3 · Cited by 15
“A network named DeepVO [27] is a recurrent CNN that calculates the camera pose from a sequence of input images in separated branches, the proposed network can process the pose with high accuracy; however, the reported results do not give hints that they can estimate in real-time.”
BackgroundMDN-VO: Estimating Visual Odometry with Confidence
Kaygusuz, N., Méndez, O., & Bowden, R. (2021) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros51168.2021.9636827 · Cited by 15
“[31], [32] propose using RNNs to model changes in image features and achieve competitive results to traditional VO based approaches.”
BackgroundMethodResultInfluentialKreuzig, R., Ochs, M., & Mester, R. (2019) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw.2019.00165 · Cited by 15
“Since DistanceNet performs better than all SRNN variations on RMSE as well as on accuracy and we declared SRNN-channel as a bound for DeepVO and ESP-VO, we probably perform better than these as well.”
MethodInfluentialImproving Learning-based Ego-motion Estimation with Homomorphism-based Losses and Drift Correction
Wang, X., Maturana, D., Yang, S., Wang, W., Chen, Q., & Scherer, S. (2019) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros40897.2019.8968515 · Cited by 15
“We first introduce the commonly used L2 loss [6]”
MethodInfluentialMiniVO: Minimalistic Range Enhanced Monocular System for Scale Correct Pose Estimation
Giubilato, R., Chiodini, S., Pertile, M., & Debei, S. (2020) · IEEE Sensors Journal · doi:10.1109/jsen.2020.2978334 · Cited by 15
“Depth predicted by deep neural networks have been integrated in Visual SLAM [25], [26].”
MethodAn Overview of Perception and Decision-Making in Autonomous Systems in the Era of LearningPreprint
Tang, Y. S., Zhao, C., Wang, J., Zhang, C., Sun, Q., Zheng, W. X., Du, W., Qian, F., & Kurths, J. (2020) · arXiv (Cornell University) · Cited by 15
RPNet: Rotational pooling net for efficient Micro Aerial Vehicle trail navigation
Agyemang, I. O., Zhang, X., Acheampong, D., Adjei-Mensah, I., Gyamfi, E. O., Arhin, J. R., Ayivi, W., & Chikwendu, I. A. (2022) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2022.105468 · Cited by 14
CCVO: Cascaded CNNs for Fast Monocular Visual Odometry Towards the Dynamic Environment
Zhang, T., Li, N., Gong, G., Yang, C., Hou, G., & Lin, X. (2022) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2022.3214790 · Cited by 14
“[17] introduced sequential learning to pose estimation using deep Recurrent Convolutional Neural Networks.”
MethodUnsupervised Visual Odometry and Action Integration for PointGoal Navigation in Indoor Environment
Cao, Y., Zhang, X.-S., Luo, F., Lin, C., & Li, Y. (2023) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2023.3263484 · Cited by 16
“The supervised meth-ods minimize the distance between predicted values (depth and ego-motion) and corresponding ground truth by using such strategies as a recurrent neural network (RNN) [34], memory mechanism [35], or feature-metric BA [36].”
MethodLightweight, Uncertainty-Aware Conformalized Visual Odometry
Stutts, A. C., Erricolo, D., Tulabandhula, T., & Trivedi, A. R. (2023) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros55552.2023.10341924 · Cited by 14
“DeepVO [14] demonstrated the combined strength of recurrent convolutional neural networks (RCNNs) in learning features and modeling sequential relationships between consecutive camera frames.”
BackgroundGlobal-Context-Aware Visual Odometry System With Epipolar-Geometry-Constrained Loss Function
Ji, Z., Nan, K., Xu, Y., Wang, H., Li, X., & Bai, J. (2024) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3370804 · Cited by 14
“DeepVO [9] leverages convolutional neural network (CNN) for feature extraction and recurrent neural network (RNN) for sequential learning to predict 6-DoF relative poses.”
BackgroundMethodInfluentialMS-VRO: A Multistage Visual-Millimeter Wave Radar Fusion Odometry
Cheng, Y., Jiang, M., & Liu, Y. (2024) · IEEE Transactions on Robotics · doi:10.1109/tro.2024.3400941 · Cited by 14
Imperative learning: A self-supervised neuro-symbolic learning framework for robot autonomy
Wang, C., Ji, K., Geng, J., Ren, Z., Fu, T., Yang, F., Guo, Y., He, H., Chen, X., Zhan, Z., et al. (2025) · The International Journal of Robotics Research · doi:10.1177/02783649251353181 · Cited by 29
“5 ◦ of its final steady attitude, measuring how quickly the system settles; RMSE is the root-mean-square of the difference between estimated and the desired attitude; and SSE is the absolute difference between the steady and desired attitude, evaluating the control accuracy.”
Three-Dimensional Dense Reconstruction: A Review of Algorithms and Datasets
Lee, Y. (2024) · Sensors · doi:10.3390/s24185861 · Cited by 14
“Deep Visual Odometry (DeepVO) estimates camera motion and depth from video sequences by integrating CNNs and RNNs [52].”
BackgroundReinforcement Learning Meets Visual Odometry
Messikommer, N., Cioffi, G., Gehrig, M., & Scaramuzza, D. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-73202-7_5 · Cited by 14
Advanced Camera Image Cropping Approach for CNN-Based End-to-End Controls on Sustainable Computing
Sung, Y., Jin, Y., Kwak, J., Lee, S.-G., & Cho, K. (2018) · Sustainability · doi:10.3390/su10030816 · Cited by 14
“This section describes the proposed method for cropping images captured by a camera mounted on a car to enhance learning performance for self-driving based on the angles of the steering wheel and the cropped images.”
MethodLong-Term Knowledge Distillation of Visual Place Classifiers
Tomoe, H., & Kanji, T. (2019) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc.2019.8917199 · Cited by 14
“, using pose tracking [51] and relocation [52]), to collect mapped images that have global viewpoint information, and optionally, the collected data may be further post-processed to refine the viewpoint information by structure-from-motion (SfM) [53] or SLAM [54].”
MethodKanayama, H., Ueda, T., Ito, H., & Yamamoto, K. (2020) · IEEE/SICE International Symposium on System Integration (SII) · doi:10.1109/sii46433.2020.9025851 · Cited by 14
“Therefore, the P-DNN is newly designed for high accuracy estimation inspired by the network for optical flow estimation in [25] and visual odometry in [19].”
BackgroundMethodCross-Modal Knowledge Distillation for Depth Privileged Monocular Visual Odometry
Li, B., Wang, S., Ye, H., Gong, X., & Xiang, Z. (2022) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2022.3166457 · Cited by 14
Self-citationAn LSTM Network for Real-Time Odometry Estimation
Valente, M., Joly, C., & Fortelle, A. D. L. (2019) · IEEE Intelligent Vehicles Symposium · doi:10.1109/ivs.2019.8814133 · Cited by 14
“Visual Odometry (VO) methods, such as DeepVO [15], that not only extracts the features of the input data but also estimate the possible connections among consecutive image inputs.”
MethodInfluentialProximity Estimation Using Vision Features Computed On Sensor
Chen, J., Liu, Y., Carey, S. J., & Dudek, P. (2020) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra40945.2020.9197370 · Cited by 14
“Convolutional neural network (CNN) based methods can provide end-to-end solutions when the computational power allowed on-board the vehicle is sufficient [1].”
BackgroundA Deep Learning Approach to Localization for Navigation on a Miniature Autonomous Blimp
Seguin, L., Zheng, J., Li, A. Q., Tao, Q., & Zhang, F. (2020) · International Conference on Control and Automation · doi:10.1109/icca51439.2020.9264514 · Cited by 14
“In [16], a recurrent convolutional neural network (RCNN) is used to learn monocular VO.”
MethodBotVIO: A Lightweight Transformer-Based Visual–Inertial Odometry for Robotics
Wei, W., Zhou, Y., Hu, Y., Li, Z., Wang, S., Liu, X., & Li, J. (2025) · IEEE Transactions on Robotics · doi:10.1109/tro.2025.3577054 · Cited by 14
Self-citationCalibBD: Extrinsic Calibration of the LiDAR and Camera Using a Bidirectional Neural Network
Nguyen, A. D., & Yoo, M. (2022) · IEEE Access · doi:10.1109/access.2022.3222797 · Cited by 13
Amarasinghe, C. K., Rathnaweera, A., & Maithripala, S. (2023) · Arabian Journal for Science and Engineering · doi:10.1007/s13369-023-07906-6 · Cited by 13
“There is very limited work on VO, even related to 3D geometry problems, and none of them address the underwater VO specifically [54].”
Muñoz, B., & Troni, G. (2024) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2024.3352357 · Cited by 13
“For instance, DeepVO [19] proposes a RCNN (a combination of a CNN with a RNN) to use sequential information in motion estimation.”
BackgroundMethodInfluentialMerveille, F. F. R., Jia, B., Xu, Z., & Fred, B. (2024) · Sensors · doi:10.3390/s24217034 · Cited by 13
From Local Understanding to Global Regression in Monocular Visual Odometry
Esfahani, M. A., Wu, K., Yuan, S., & Wang, H. (2019) · International Journal of Pattern Recognition and Artificial Intelligence · doi:10.1142/s0218001420550022 · Cited by 13
Unsupervised Deep Learning-Based RGB-D Visual Odometry
Liu, Q., Zhang, H., Xu, Y., & Wang, L. (2020) · Applied Sciences · doi:10.3390/app10165426 · Cited by 13
“[34] further extended the CNN-based PoseNet by incorporating a Recurrent Neural Network (RNN) into their system.”
BackgroundMethodTwo-stage visual navigation by deep neural networks and multi-goal reinforcement learning
Shantia, A., Timmers, R., Chong, Y., Kuiper, C., Bidoia, F., Schomaker, L., & Wiering, M. (2021) · Robotics and Autonomous Systems · doi:10.1016/j.robot.2021.103731 · Cited by 14
“[41], proposed a deep visual odometry method using a recurrent convolutional neural network architecture that receives a pair of images in each forward pass and outputs the poses.”
MethodStereo Visual Odometry Pose Correction through Unsupervised Deep Learning
Zhang, S., Lu, S., He, R., & Bao, Z. (2021) · Sensors · doi:10.3390/s21144735 · Cited by 13
“[28] proposed the DeepVO, which utilizes a combination of CNN and RNN to estimate directly poses from the raw RGB image.”
MethodContextAVO: Local context guided and refining poses for deep visual odometry
Song, R., Zhu, R., Xiao, Z., & Yan, B. (2023) · Neurocomputing · doi:10.1016/j.neucom.2023.02.014 · Cited by 13
Deep Radio-Visual Localization
Ishihara, T., Kitani, K., Asakawa, C., & Hirose, M. (2018) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv.2018.00071 · Cited by 13
“This has also been applied in other areas, such as camera relocalization [23], relative camera motion [28], visual odometory [40], and RANSAC pose estimation [8].”
MethodImproving drone localisation around wind turbines using monocular model-based tracking
Moolan-Feroze, O., Karachalios, K., Nikolaidis, D., & Calway, A. D. (2019) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2019.8794156 · Cited by 13
“This method is also used in [10], where the authors state the recurrent layers are able to learn the motion dynamics of the system over a period of time.”
Singh, A., Narula, R., Rashwan, H. A., Abdel‐Nasser, M., Puig, D., & Nandi, G. C. (2022) · Neural Computing and Applications · doi:10.1007/s00521-022-07273-7 · Cited by 13
“Furthermore, deep learning algorithms have been shown to be effective in the learning of visual odometry [7, 8].”
BackgroundMethodResultInfluentialGao, F., Yu, J., Shen, H., Wang, Y., & Yang, H. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2011.09369 · Cited by 15
“Recently, learning-based depth estimation [2, 3, 4, 5] and ego-motion estimation [6, 7, 8, 9] methods have emerged and achieved great success.”
DeepSFM: Robust Deep Iterative Refinement for Structure From Motion
Ren, X., Wei, X., Li, Z., Fu, Y., Zhang, Y., & Xue, X. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3307567 · Cited by 15
Cai, Y., Qin, T., Ou, Y., & Wei, R. (2023) · International Journal on Semantic Web and Information Systems · doi:10.4018/ijswis.333056 · Cited by 12
“…monocular SLAM with learned depth prediction (Tateno et al., 2017) DeepVO: Towards end-to-end visual odometry with deep recurrent convolutional neural networks (S. Wang, Clark, et al., 2017) algorithms like SE-Sync by Dong et al. (2015) offered a certifiable solution to the Graph SLAM problem.”
MethodZhou, Y., Li, X., Li, S., Wang, X., Feng, S., & Tan, Y. (2024) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2024.3400156 · Cited by 17
“There are also many works that attempt to develop end-to-end visual SLAM systems [12], [13].”
BackgroundMonocular Visual Odometry using Learned Repeatability and Description
Huang, H., Ye, H., Sun, Y., & Liu, M. (2020) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra40945.2020.9197406 · Cited by 12
“While some existing works generally recovering pose in an end-to-end manner [16], [17], limited the generalization ability for different scenarios, we consider a more practical approach to utilize the deep feature.”
MethodRobustness Improvement of Using Pre-Trained Network in Visual Odometry for On-Road Driving
Chen, W., Zhu, L., Loo, S. Y., Wang, J., Wang, C., Meng, M. Q. ‐., & Zhang, H. (2021) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2021.3120214 · Cited by 12
Álvarez-Tuñón, O., Brodskiy, Y., & Kayacan, E. (2023) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2023.3321032 · Cited by 15
“Supervised pose regression networks usually rely on a Euclidean loss for rotation and translation [127], [129], [136], [137].”
BackgroundMethodFlow-based SLAM: From geometry computation to learning
Yan, Z., & Zha, H. (2019) · Virtual Reality & Intelligent Hardware · doi:10.1016/j.vrih.2019.09.001 · Cited by 11
“Due to the computationally expensive optical flow estimation, the encoder of Flownet [102] is taken for spatio-temporal feature extraction [101, 103, 106, 108] .”
BackgroundMethodOverfitting reduction of pose estimation for deep learning visual odometry
Yang, X., Li, X., Guan, Y., Song, J., & Wang, R. (2020) · China Communications · doi:10.23919/jcc.2020.06.016 · Cited by 12
Metrically-Scaled Monocular SLAM using Learned Scale Factors
Greene, W. N., & Roy, N. (2020) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra40945.2020.9196900 · Cited by 12
“To address these limitations, methods that apply end-to-end deep learning techniques to the monocular SLAM problem have appeared in recent years, spurred by the rapid adoption of deep, convolutional neural networks (CNNs) for a variety of computer vision tasks [14]–[18].”
BackgroundMethodLighting Enhancement Aids Reconstruction of Colonoscopic Surfaces
Zhang, Y., Wang, S., Ma, R., McGill, S., Rosenman, J. G., & Pizer, S. M. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-78191-0_43 · Cited by 12
Self-citationA Survey of Simultaneous Localization and Mapping with an Envision in 6G Wireless Networks
Huang, B., Zhao, J., Luo, S., & Liu, J. (2021) · Journal of Global Positioning Systems · doi:10.5081/jgps.17.1.94 · Cited by 59
WiFi Similarity-Based Odometry
Ismail, K., Liu, R., Athukorala, A., Ng, B. K. K., Yuen, C., & Tan, U.-X. (2023) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2023.3274924 · Cited by 11
“Visual odometry methods, that are learning-based, apply machine learning techniques to estimate odometry by using a model that has been trained on labeled data sets [11], [12], [13].”
MethodMinimal configuration point cloud odometry and mapping
Bhandari, V., Phillips, T. G., & McAree, P. R. (2024) · The International Journal of Robotics Research · doi:10.1177/02783649241235325 · Cited by 11
“Considerable research has been conducted into deep learning frameworks by Wang et al. (2017), Li et al. (2019), Li and Wang (2020), and Zheng et al. (2020).”
UVMO: Deep unsupervised visual reconstruction-based multimodal-assisted odometry
Han, S., Li, M., Tang, H., Song, Y., & Tong, G. (2024) · Pattern Recognition · doi:10.1016/j.patcog.2024.110573 · Cited by 11
CeMNet: Self-Supervised Learning for Accurate Continuous Ego-Motion Estimation
Lee, M., & Fowlkes, C. C. (2019) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw.2019.00048 · Cited by 11
“[38] train a recurrent CNN to capture longer-term processing of sequences typically handled by bundle adjustment and loop closure.”
BackgroundDynamic Attention-based Visual Odometry
Kuo, X.-Y., Liu, C., Lin, K.-C., Luo, E., Chen, Y., & Lee, C. (2020) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros45743.2020.9340890 · Cited by 11
“As flow maps contain rich information relevant to the motion of the camera [4], [10-12] as well as those of the perceived semantic contents, it is reasonable to distinguish different motion scenarios by leveraging on them.”
BackgroundMethodInfluentialMun, J.-H., Jeon, M., & Lee, B. (2019) · Sensors · doi:10.3390/s19112459 · Cited by 11
Tanaka, T., Sasagawa, Y., & Okatani, T. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00619 · Cited by 11
“To improve the accuracy and robustness of 3D reconstruction, researchers have considered several different approaches, such as feature-point-based methods [27], direct methods [7], learning-based methods [24, 23], and their hybrids [3].”
Multi-Camera Sensor Fusion for Visual Odometry using Deep Uncertainty Estimation
Kaygusuz, N., Méndez, O., & Bowden, R. (2021) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc48978.2021.9565079 · Cited by 15
“More recently, deep learning based models have been applied to the VO task [10], [11], [12].”
COMPASS: Contrastive Multimodal Pretraining for Autonomous Systems
Ma, S., Vemprala, S. H., Wang, W., Gupta, J. K., Song, Y., McDufft, D., & Kapoor, A. (2022) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros47612.2022.9982241 · Cited by 12
“al [53], TartanVO [50]) and unsupervised methods (UnDeepVO [55], GeoNet [56]).”
MethodH-VLO: Hybrid LiDAR-Camera Fusion For Self-Supervised Odometry
Aydemir, E., Fetic, N., & Ünel, M. (2022) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros47612.2022.9981111 · Cited by 11
“Table I provides a comparison between our and the state-of-the-art algorithms, as reported in the literature, such as classical VO [4], unsupervised VOs [20, 15] supervised VO [19], supervised LO [10], unsupervised LO [9], hybrid VOs [7, 21] and lastly learning-based VLO [3].”
MethodNeurAll: Towards a Unified Model for Visual Perception in Automated DrivingPreprint
Sistu, G., Leang, I., Chennupati, S., Hughes, C., Milz, S., Yogamani, S. K., & Rawashdeh, S. A. (2019) · arXiv (Cornell University) · Cited by 11
MaskVO: Self-Supervised Visual Odometry with a Learnable Dynamic Mask
Xuan, W., Ren, R., Wu, S., & Chen, C. (2022) · IEEE/SICE International Symposium on System Integration (SII) · doi:10.1109/sii52469.2022.9708796 · Cited by 11
EffLoc: Lightweight Vision Transformer for Efficient 6-DOF Camera Relocalization
Xiao, Z., Chen, C., Yang, S., & Wu, W. (2024) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra57147.2024.10611622 · Cited by 15
“Other end-to-end leaning approaches such as [8] utilize an implicit map database to store scene information and eliminate the complex handcrafted feature engineering.”
Visual-Inertial-Wheel Odometry with Slip Compensation and Dynamic Feature Elimination
Reginald, N., Al-Buraiki, O., Choopojcharoen, T., Fi̇dan, B., & Hashemi, E. (2025) · Sensors · doi:10.3390/s25051537 · Cited by 14
“Similarly, deep-learning-based visual odometry (VO) methods, such as recurrent convolutional neural networks (RCNNs) and bi-directional LSTM (Bi-LSTM) architectures, have been used to improve pose estimation accuracy by capturing both spatial and temporal dependencies [29,30].”
ForestVO: Enhancing Visual Odometry in Forest Environments Through ForestGlue
Pritchard, T., Ijaz, S., Clark, R., & Koçer, B. B. (2025) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2025.3557738 · Cited by 10
“Furthermore, deep learning in VO systems [20] has been explored as an alternative end-to-end approach, directly regressing the relative camera poses from the input images, instead of geometric-based pose estimation methods.”
Self-citationA New Approach to Train Convolutional Neural Networks for Real-Time 6-DOF Camera Relocalization
Esfahani, M. A., Wu, K., Yuan, S., & Wang, H. (2018) · International Conference on Control and Automation · doi:10.1109/icca.2018.8444299 · Cited by 10
Deep Bayesian ICP Covariance Estimation
Maio, A. D., & Lacroix, S. (2022) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra46639.2022.9811899 · Cited by 10
Deep Feature-based RGB-D Odometry using SuperPoint and SuperGlue
Fujimoto, S., & Matsunaga, N. (2023) · Procedia Computer Science · doi:10.1016/j.procs.2023.10.624 · Cited by 10
Ott, F., Heublein, L., Rügamer, D., Bischl, B., & Mutschler, C. (2024) · Journal of Visual Communication and Image Representation · doi:10.1016/j.jvcir.2024.104256 · Cited by 12
“Relative pose regression (RPR)methods (Wang et al., 2017; Iyer et al., 2018; Kreuzig et al., 2019; Idan et al., 2023) have emerged, which predict the relative pose by using convolutional neural networks (CNNs) and recurrent neural networks (RNNs).”
BackgroundMethodA Combined RGB and Depth Descriptor for SLAM with Humanoids
Sheikh, R., Obwald, S., & Bennewitz, M. (2018) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2018.8593768 · Cited by 10
Unsupervised Ego-Motion and Dense Depth Estimation with Monocular Video
Xu, Y., Wang, Y., & Guo, L. (2018) · IEEE International Conference on Communication Technology (ICCT) · doi:10.1109/icct.2018.8600039 · Cited by 11
MODELLING UNCERTAINTY OF SINGLE IMAGE INDOOR LOCALISATION USING A 3D MODEL AND DEEP LEARNING
Acharya, D., Roy, S. S., Khoshelham, K., & Winter, S. (2019) · ISPRS annals of the photogrammetry, remote sensing and spatial information sciences · doi:10.5194/isprs-annals-iv-2-w5-247-2019 · Cited by 10
“, 2018) and visual odometry (Wang et al., 2017) use a sequence of images, and direct pose regression approaches, such as PoseNet (Kendall et al.”
MethodEstimation of Absolute Scale in Monocular SLAM Using Synthetic Data
Rukhovich, D., Mouritzen, D., Kaestner, R., Rufli, M., & Velizhev, A. (2019) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw.2019.00108 · Cited by 11
“For example, the DeepVO method [32] introduces a fully connected layer with 122M trainable parameters and thus significantly increases computational complexity with strong implications to real-time applications.”
MethodInfluentialMobile Robot Self-LocalizationUsing Visual Odometry Based on Ceiling Vision
Lin, Q., Liu, X., & Zhang, Z. (2019) · IEEE Symposium Series on Computational Intelligence (SSCI) · doi:10.1109/ssci44817.2019.9003092 · Cited by 10
“In previous works, most of the visual odometry have used front view [3]–[6].”
BackgroundTowards Video Streaming Analysis and Sharing for Multi-Device Interaction with Lightweight DNNs
Huang, Y., Zhao, H., Qiao, X., Tang, J., & Liu, L. (2021) · IEEE Conference on Computer Communications (INFOCOM) · doi:10.1109/infocom42981.2021.9488846 · Cited by 10
“However, heavy model size and intensive inference hinder us from using them for implementing real-time pose estimation via VO streaming (e.g., DeepVO [25], whose model size is 268MB, requires deploying GPUs in the cloud to support such video streaming processing).”
BackgroundMethodInfluentialA Survey of Simultaneous Localization and Mapping with an Envision in 6G Wireless Networks
Huang, B., Zhao, J., Luo, S., & Liu, J. (2021) · Journal of Global Positioning Systems · doi:10.5081/jgps.17.2.206 · Cited by 11
Deep Learning Localization with 2D Range Scanner
Spampinato, G., Bruna, A. R., Guarneri, I., & Giacalone, D. (2021) · International Conference on Automation, Robotics and Applications · doi:10.1109/icara51699.2021.9376424 · Cited by 10
“processed 3D laser scanner data in CNNs to estimate odometry [15], Pfeiffer et al.”
MethodPerspectives on Deep Multimodel Robot Learning
Burgard, W., Valada, A., Radwan, N., Naseer, T., Zhang, J., Vertens, J., Mees, O., Eitel, A., & Oliveira, G. L. (2019) · Springer proceedings in advanced robotics · doi:10.1007/978-3-030-28619-4_3 · Cited by 10
“In [29], the authors use a AlexNet-based Siamese architecture and an L2-loss layer with equal weights for the translational and rotational components, while in [15] the authors propose a weighting term to balance these components.”
MethodJiao, J., Jiao, J., Mo, Y., Liu, W., & Deng, Z. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.10964 · Cited by 10
Probabilistic Regression of Rotations using Quaternion Averaging and a Deep Multi-Headed NetworkPreprint
Peretroukhin, V., Wagstaff, B., Giamou, M., & Kelly, J. W. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1904.03182 · Cited by 10
“7 km) DeepVO [30] — — — — SfMLearner [32] — — 65.”
4DEgo: ego-velocity estimation from high-resolution radar data
Kumar, P., Strokina, N., & Ghabcheloo, R. (2023) · Frontiers in Signal Processing · doi:10.3389/frsip.2023.1198205 · Cited by 9
“A fully connected regressor network can perform descriptor (feature vector) matching by finding the closest match from one heatmap to the other in the pair based on geometry (Euclidean distance in our case) (Wang et al., 2017) (Costante et al.”
BackgroundModality-invariant Visual Odometry for Embodied Vision
Memmel, M., Bachmann, R., & Zamir, A. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.02064 · Cited by 9
“Method Observations Samples S ↑ SPL↑ SSPL↑ dg ↓ oracle GPS+Compass 250 k 97 74 73 29 VOT-B (MultiMAE) RGB 250 k 59 45 66 66 VOT-B (MultiMAE) Depth 250 k 93 71 72 38 VOT-B (MultiMAE) RGB-D 250 k 88 67 71 42 ResNet-50, unified [12] RGB 250 k 45 34 63 95 ResNet-50, unified [12] Depth 250 k 59 45 65 81 ResNet-50, unified…”
TEFu-Net: A time-aware late fusion architecture for robust multi-modal ego-motion estimation
Agostinho, L. R., Pereira, D., Hiolle, A., & Pinto, A. M. (2024) · Robotics and Autonomous Systems · doi:10.1016/j.robot.2024.104700 · Cited by 8
GraphAVO: Self-Supervised Visual Odometry Based on Graph-Assisted Geometric Consistency
Song, R., Liu, J., Xiao, Z., & Yan, B. (2024) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2024.3462596 · Cited by 9
Spatiotemporal Dual-Stream Network for Visual Odometry
Xu, C., Zeng, T., Luo, Y., Song, F., & Si, B. (2025) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2025.3544521 · Cited by 9
Temporal Convolutions for Multi-Step Quadrotor Motion Prediction
Looper, S., & Waslander, S. L. (2022) · Canadian Conference on Computer and Robot Vision · doi:10.1109/crv55824.2022.00013 · Cited by 9
Visual Odometry in Challenging Environments: An Urban Underground Railway Scenario Case
Etxeberria-Garcia, M., Zamalloa, M., Arana-Arexolaleiba, N., & Labayen, M. (2022) · IEEE Access · doi:10.1109/access.2022.3187209 · Cited by 9
“Most state-of-the-art VO approaches are evaluated in the standard KITTI [2, 3] vision benchmark [36, 82, 42, 29, 17, 43].”
BackgroundMethodUnsupervised Learning of Accurate Camera Pose and Depth From Video Sequences With Kalman Filter
Wang, Y., & Xu, Y.-F. (2019) · IEEE Access · doi:10.1109/access.2019.2903871 · Cited by 9
“All of these works build encoder-decoder networks based on VGG or ResNet [6], [7], [17], [18], with the deconvolution as an upsampling model.”
MethodEstimating Metric Scale Visual Odometry from Videos using 3D Convolutional Networks
Koumis, A. S., Preiss, J. A., & Sukhatme, G. S. (2019) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros40897.2019.8967919 · Cited by 10
InertialNet: Toward Robust SLAM via Visual Inertial Measurement
Liu, T.-A., Lin, H., & Lin, W.-Y. (2019) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc.2019.8917003 · Cited by 9
“In this paper, we adopt the optical flow backbone design of DeepVO for generalization purpose.”
BackgroundMethodInfluentialA Review of Visual Odometry in SLAM Techniques
Yang, G., Wang, Y., Zhi, J., Liu, W., Shao, Y., & Peng, P. (2020) · International Conference on Artificial Intelligence and Electromechanical Automation (AIEA) · doi:10.1109/aiea51086.2020.00075 · Cited by 9
“During training, the relative poses of multiple cameras are used to predict the appearance of nearby images, and then the most appropriate depth is selected based on the amount of plane scans to sample colors from adjacent images[17-18].”
Unsupervised Deep Persistent Monocular Visual Odometry and Depth Estimation in Extreme Environments
Almalıoğlu, Y., Santamaria‐Navarro, À., Morrell, B. J., & Agha–mohammadi, A. (2021) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros51168.2021.9636555 · Cited by 9
Scene-Aware Error Modeling of LiDAR/Visual Odometry for Fusion-Based Vehicle Localization
Ju, X., Xu, D., & Zhao, H. (2021) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2021.3058054 · Cited by 10
Zhao, B., Huang, Y., Ci, W., & Hu, X. (2022) · Sensors · doi:10.3390/s22041383 · Cited by 9
“DeepVO [10] is a typical supervised learning method to estimate camera motion.”
Method
Metadata
- Authors
- Wang, S., Clark, R., Wen, H., & Trigoni, N.
- Affiliations
- University of Oxford (GB)
- Venue
- 2017 IEEE International Conference on Robotics and Automation (ICRA)
- Publisher
- IEEE
- Published
- 2017-05
- Pages
- 2043-2050
- ISSN
- 1050-4729 (ISSN-L)
- 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
- green
- arxiv.org
- 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 › 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
- 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
