The DOI matches its Crossref record; 30 of the 30 references it cites were found in a registry or an index.
Modelling uncertainty in deep learning for camera relocalization
Open accessAbstract
We present a robust and real-time monocular six degree of freedom visual relocalization system. We use a Bayesian convolutional neural network to regress the 6-DOF camera pose from a single RGB image. It is trained in an end-to-end manner with no need of additional engineering or graph optimisation. The algorithm can operate indoors and outdoors in real time, taking under 6ms to compute. It obtains approximately 2m and 6° accuracy for very large scale outdoor scenes and 0.5m and 10° accuracy indoors. Using a Bayesian convolutional neural network implementation we obtain an estimate of the model's relocalization uncertainty and improve state of the art localization accuracy on a large scale outdoor dataset. We leverage the uncertainty measure to estimate metric relocalization error and to detect the presence or absence of the scene in the input image. We show that the model's uncertainty is caused by images being dissimilar to the training dataset in either pose or appearance.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
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?
In the citing paper
“We adopted the ResNet architectures with pre-trained weights as the network backbone, due to its low number of pooling layers, and good accuracy-complexity trade-off [35].”
From Deep Learning for Spacecraft Pose Estimation from Photorealistic Rendering (2020) · cites this paper as [35]
Citations
Cites
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 1×
“Convolutional neural networks have recently been shown to be extremely powerful for generating image representations for classification [12], [13].”
BackgroundDistinctive Image Features from Scale-Invariant Keypoints
Lowe, D. G. (2004) · International Journal of Computer Vision · doi:10.1023/b:visi.0000029664.99615.94 · Cited by 55,589 · cited 3×
“This has been achieved using SIFT [2] features with FAB-MAP [6] or using convolutional neural network features [7].”
BackgroundMethodGoing 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 · cited 4×
“Additionally, in green we show the output from the first auxiliary pose regressor from the GoogLeNet architecture (see Figure 3 of [13]).”
BackgroundMethodInfluential(2014) · Cited by 33,419
Visualizing and understanding convolutional networks
Zeiler, M. D., & Fergus, R. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-10590-1_53 · Cited by 15,839 · cited 1×
“These higher level contours and shape cues [14] have been shown to be increasingly invariant to object instantiation parameters.”
BackgroundCaffe: Convolutional Architecture for Fast Feature Embedding
Jia, Y., Shelhamer, E., Donahue, J., Karayev, S., Long, J., Girshick, R., Guadarrama, S., & Darrell, T. (2014) · Proceedings of the 22nd ACM international conference on Multimedia · doi:10.1145/2647868.2654889 · Cited by 11,077 · cited 1×
“Using standard libraries such as [25] we can obtain an efficient implementation of a Bernoulli Bayesian convolutional neural network.”
MethodORB: An efficient alternative to SIFT or SURF
Rublee, E., Rabaud, V., Konolige, K., & Bradski, G. (2011) · 2011 International Conference on Computer Vision · doi:10.1109/iccv.2011.6126544 · Cited by 11,037 · cited 2×
“Visual SLAM [1], [4], [5], [9], and relocalization research [10], [11], has largely focused on registering viewpoints using point-based landmark features, such as SIFT [2] or ORB [3].”
BackgroundMethodThrun, S. (2002) · CACM · doi:10.1145/504729.504754 · Cited by 7,941 · cited 1×
“nsider localization uncertainty [1], [5], [11], previous probabilistic algorithms have been proposed. Bayesian approaches include extended Kalman filters and particle filter approaches such as FastSLAM [18]. However these approaches estimate uncertainty from sensor noise models, not the uncertainty of the model to repre…”
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 1×
“Visual SLAM [1], [4], [5], [9], and relocalization research [10], [11], has largely focused on registering viewpoints using point-based landmark features, such as SIFT [2] or ORB [3].”
BackgroundDropout as a bayesian approximation: Representing model uncertainty in deep learning
Gal, Y., & Ghahramani, Z. (2015) · arXiv · doi:10.48550/arxiv.1506.02142 · Cited by 6,471 · cited 1×
“Recent work [23] has extended dropout to approximate Bayesian inference over the network’s weights.”
Background
Flagged first, then by citations
Cited by
Sorted by citations
Items 301–400 of 549
Unsupervised Learning of Camera Pose with Compositional Re-estimation
Nabavi, S. S., Hosseinzadeh, M., Fahimi, R., & Wang, Y. (2020) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv45572.2020.9093495 · Cited by 7
“Recently, learningbased methods have become popular and raised the bar on the performance [40, 15, 14, 27].”
BackgroundCan You Trust Your Pose? Confidence Estimation in Visual Localization
Ferranti, L., Li, X., Boutellier, J. J., & Kannala, J. (2021) · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · doi:10.1109/icpr48806.2021.9413234 · Cited by 7
“proposed in [12] a Bayesian pose regression neural network which, instead of producing a point prediction of the pose, produced a confidence interval for each scalar parameter.”
BackgroundA Structure-Aware Method for Direct Pose Estimation
Blanton, H., Workman, S., & Jacobs, N. (2022) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv51458.2022.00028 · Cited by 7
“Many works have explored the application of CNNs to this problem, but differences lie largely in changes to the underlying feature extraction architecture or modified objective functions [21, 22, 29, 42].”
BackgroundKuo, Y.-L., & Tang, S.-C. (2021) · Transactions of the Institute of Measurement and Control · doi:10.1177/01423312211002795 · Cited by 7
“Kendall and Cipolla (2016) used a Bayesian convolutional neural network to develop a visual relocalization system, which can estimate the relocalization uncertainty and improve the localization accuracy.”
MethodLiu, Y., Zhao, C., & Wei, Y. (2022) · Applied Sciences · doi:10.3390/app12063007 · Cited by 7
Deep 6-DoF camera relocalization in variable and dynamic scenes by multitask learning
Wang, J., & Qi, Y. (2023) · Machine Vision and Applications · doi:10.1007/s00138-023-01388-0 · Cited by 7
“Direct learning-based algorithms regress the camera pose of an input image by training CNNs, which employ network parameters to represent the scene [1,7,22–24, 27 ,33, 52].”
MethodDNN Based Speech Enhancement for Unseen Noises Using Monte Carlo Dropout
Nazreen, P. M., & Ramakrishnan, A. G. (2018) · International Conference on Signal Processing and Communication Systems · doi:10.1109/icspcs.2018.8631773 · Cited by 7
“The uncertainty is computed by taking the trace of the covariance matrix of each frame [25].”
BackgroundMethodResultInfluentialAccurate Sparse Feature Regression Forest Learning for Real-Time Camera Relocalization
Duong, N.-D., Kacete, A., Soladié, C., Richard, P.-Y., & Royan, J. (2018) · International Conference on 3D Vision (3DV) · doi:10.1109/3dv.2018.00079 · Cited by 7
“By performing a dropout, [14] generates a probabilistic pose estimation.”
BackgroundMethodAccurate Deep Direct Geo-Localization from Ground Imagery and Phone-Grade GPS
Sun, S., Sarukkai, R. R., Kwok, J., & Shet, V. D. (2018) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw.2018.00148 · Cited by 7
“In [5], an extension to PoseNet evaluates the CNN with a fraction of its neurons randomly disabled.”
MethodDuong, N.-D., Kacete, A., Sodalie, C., Richard, P.-Y., & Royan, J. (2018) · IEEE International Symposium on Mixed and Augmented Reality Adjunct (ISMAR-Adjunct) · doi:10.1109/ismar-adjunct.2018.00080 · Cited by 7
“The deep learning based methods in [5, 15, 17, 28] estimate directly camera pose from whole images.”
MethodMonocular Object Orientation Estimation using Riemannian Regression and Classification NetworksPreprint
Mahendran, S., Lu, M. Y., Ali, H., & Vidal, R. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1807.07226 · Cited by 9
“There has been recent interest in quaternion representations of 3D pose for camera localization (Kendall et al, 2015; Kendall and Cipolla, 2016, 2017; Tulsiani et al, 2018), but these works recommend a Euclidean, repro-jection or classification loss respectively instead of the geodesic loss.”
BackgroundHGSLoc: 3DGS-Based Heuristic Camera Pose Refinement
Niu, Z., Tan, Z., Zhang, J., Yang, X., & Hu, D. (2025) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra55743.2025.11127431 · Cited by 6
Deep Learning For Pose Estimation From Event Camera
Tabia, A., Bonardi, F., & Bouchafa, S. (2022) · International Conference on Digital Image Computing: Techniques and Applications (DICTA) · doi:10.1109/dicta56598.2022.10034617 · Cited by 6
“We use the same protocol of comparison reported in [27] and used in PoseNet [14] and Bayesian PoseNet [12].”
BackgroundMethodInfluentialLamping, C., Kootstra, G., & Derks, M. (2023) · Smart Agricultural Technology · doi:10.1016/j.atech.2023.100308 · Cited by 7
FusedNet: End-to-End Mobile Robot Relocalization in Dynamic Large-Scale Scene
Chen, F.-X., Tang, Y., Tai, C., Liu, X.-P., Wu, X., Zhang, T., & Zeng, L. (2024) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2024.3372465 · Cited by 6
“Subsequently, [8], [16] separately introduces uncertainty and learning weighted loss into the neural network to enhance robustness and strengthen network supervision.”
BackgroundVisual Localization under Appearance Change: A Filtering Approach
Doan, A.-D., Latif, Y., Chin, T.-J., Liu, Y., Chng, S.-F., Do, T.-T., & Reid, I. D. (2019) · International Conference on Digital Image Computing: Techniques and Applications (DICTA) · doi:10.1109/dicta47822.2019.8945810 · Cited by 6
“Other variants of PoseNet include: uncertainty in the pose prediction [17] and geometric constraints [10], [18].”
BackgroundA Closed-Form Uncertainty Propagation in Non-Rigid Structure From Motion
Song, J., Patel, M., Jasour, A. M. Z., & Ghaffari, M. (2022) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2022.3173733 · Cited by 6
“The uncertainty propagation of the estimated state provides important statistical descriptions for further applications such as quality assessment [5], localization [6], mapping [7], path planning [8], and multi-source information fusion [9], [10].”
BackgroundCoPR: Toward Accurate Visual Localization With Continuous Place-Descriptor Regression
Zaffar, M., Nan, L., & Kooij, J. F. P. (2023) · IEEE Transactions on Robotics · doi:10.1109/tro.2023.3262106 · Cited by 6
“Absolute Pose Regression: APR started from the seminal works of PoseNet [4] and the incremental build-up by the authors in [24] and [25] and has since seen many different variants of it e.”
MethodDeepMEL: Compiling Visual Multi-Experience Localization into a Deep Neural Network
Gridseth, M., & Barfoot, T. D. (2020) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra40945.2020.9197362 · Cited by 6
“extended the work to use a Bayesian neural network providing relocalization uncertainty [21] and an improved loss function [22].”
MethodStatistical Test Coverage for Linux-Based Next-Generation Autonomous Safety-Related Systems
Allende, I., Guire, N. M., Perez-Cerrolaza, J., Monsalve, L. G., Petersohn, J., & Obermaisser, R. (2021) · IEEE Access · doi:10.1109/access.2021.3100125 · Cited by 6
“There are also a significant number of studies that rely on probabilistic or statistical methods to advance in certain aspects of safety certification, such as Measurement-Based Probabilistic Timing Analysis (MBPTA) [37], deep-learning uncertainty [38]–[40], and scheduling [41].”
BackgroundLocal_INN: Implicit Map Representation and Localization with Invertible Neural Networks
Zang, Z., Zheng, H., Betz, J., & Mangharam, R. (2023) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48891.2023.10161015 · Cited by 6
WatchPose: A View-Aware Approach for Camera Pose Data Collection in Industrial Environments
Yang, C., Simon, G., See, J., Berger, M., & Wang, W. (2020) · Sensors · doi:10.3390/s20113045 · Cited by 7
“Specifically, Bayesian-PoseNet [39] was released by the same authors of the PoseNet approach.”
MethodInfluentialAsymptotic Prediction Error Variance for Feedforward Neural Networks
Malmström, M., Skog, I., Axehill, D., & Gustafsson, F. (2020) · IFAC-PapersOnLine · doi:10.1016/j.ifacol.2020.12.1310 · Cited by 6
“Recently, some more efficient implementations have been proposed Gustafsson et al. (2019); Blundell et al. (2015); Kendall and Cipolla (2016). proposed Gustafsson et al. (2019); Blundell et al. (2015); Kendall and Cipolla (2016).”
BackgroundMethodInfluentialLearning Multiplicative Interactions with Bayesian Neural Networks for Visual-Inertial OdometryPreprint
Shinde, K., Lee, J., Humt, M., Sezgin, A., & Triebel, R. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2007.07630 · Cited by 9
“Addition of LSTM after CNN similar to Wang et al. (2017) is made to the network in (Walch et al., 2017; Clark et al., 2017a) and the network is further improved by making the CNN part Bayesian to estimate uncertainty of localization in Kendall & Cipolla (2016).”
MethodDeep learning based monocular visual-inertial odometry
Esfahani, M. A. (2020) · doi:10.32657/10356/152280 · Cited by 6
Melvin, S. A., Wang, H., & Yazdani, A. (2026) · Robotica · doi:10.1017/s0263574725102518 · Cited by 5
“Bayesian PoseNet [58] extended this approach by leveraging uncertainty measures for estimating re-localization error and scene detection.”
Absolute Camera Pose Regression Using an RGB-D Dual-Stream Network and Handcrafted Base Poses
Kao, P.-Y., Zhang, R.-R., Chen, T., & Hung, Y. (2022) · Sensors · doi:10.3390/s22186971 · Cited by 5
“Based on the network architecture of PoseNet’s network architecture, the Bayesian PoseNet [11] adds dropout layers after FC (fully-connected) layers of position and rotation regressors.”
MethodWu, H., Cheng, C., Zhang, D., & Zhou, H. (2024) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2024.109786 · Cited by 5
Han, J., Beyer, L. L., Cavalheiro, G. V., & Karaman, S. (2024) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros58592.2024.10801527 · Cited by 5
“• We analyze the accuracy and efficiency improvements of the proposed framework compared to baselines.”
Use of LSTM Regression and Rotation Classification to Improve Camera Pose Localization Estimation
Xu, M. R., Wang, L., Ren, J., & Poslad, S. (2020) · IEEE International Conference on Anti-counterfeiting, Security, and Identification (ASID) · doi:10.1109/asid50160.2020.9271762 · Cited by 5
Malmström, M., Skog, I., Axehill, D., & Gustafsson, F. (2021) · International Conference on Information Fusion (FUSION) · doi:10.23919/fusion49465.2021.9626974 · Cited by 5
“The subject of assessing uncertainty in the prediction of NN s have recently received increased attention [20], [10], [21], [22], [23].”
BackgroundYang, S., & Ma, L. (2023) · IEEE Sensors Journal · doi:10.1109/jsen.2023.3321724 · Cited by 5
“We fine-tune PoseNet [50] and Bayesian PoseNet [51] on the original TUM-LSI dataset, and the parameter settings also refer to the experimental settings in [14].”
MethodIbrahim, M., Akhtar, N., Anwar, S., & Mian, A. S. (2023) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros55552.2023.10342046 · Cited by 5
“Currently, techniques such as PoseNet and its variants [7], [8] use a single or a series of image frames to predict 6DoF poses.”
MethodLi, D., Zhang, H., Cheng, J., & Liu, B. (2024) · Journal of Cloud Computing Advances Systems and Applications · doi:10.1186/s13677-024-00592-1 · Cited by 5
“…framework splits the network, uploads the portion that consumes computing resources to the server for calculation, and finally returns the results to the mobile terminal for subsequent processing(The network types available for both frameworks are PoseNet-new version, MapNet, and MapNet++).”
BackgroundKang, X., Liu, K., Duan, J., Gong, Y., & Qiu, G. (2024) · ISPRS Journal of Photogrammetry and Remote Sensing · doi:10.1016/j.isprsjprs.2024.04.019 · Cited by 5
Improving Image-Based Localization with Deep Learning: The Impact of the Loss Function
Ward, I. R., Jalwana, M. A. A. K., & Bennamoun, M. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-39770-8_9 · Cited by 5
“All three affine regressors’ predicBayesian Default Proposed Scene PoseNet [2] PoseNet [4] model”
BackgroundMethodGeometry-Supervised Pose Network for Accurate Retail Shelf Pose Estimation
Mou, Y., Huang, Z., Lin, L., Guo, Y., & Yang, Z. (2020) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2020.3001147 · Cited by 5
“Then, several other modifications like Hourglass-Pose [12] and BranchNet [13] based on PoseNet [20] were also proposed.”
MethodLearning Shape-based Representation for Visual Localization in Extremely Changing Conditions
Jeon, H., Im, S., Oh, J., & Hebert, M. (2020) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra40945.2020.9196842 · Cited by 5
“This work is extended by adopting a stochastic Dropout [19] in [9], and a geometric loss minimizing reprojection errors in [20].”
BackgroundMethodInfluentialVisual-based Global Localization from Ceiling Images using Convolutional Neural Networks
Scales, P., Rimel, M., & Aycard, O. (2021) · International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP) · doi:10.5220/0010248409270934 · Cited by 5
Sports Field Recognition Using Deep Multi-task Learning
Tarashima, S. (2021) · Journal of Information Processing · doi:10.2197/ipsjjip.29.328 · Cited by 5
“Specifically, recent camera pose estimation meth-ods [6], [7], [8], [9], [10], [11] fine-tune pre-trained CNN (e.g., GoogLeNet [23], ResNet [24]) to directly regress pose parameters from the input frame.”
BackgroundMethodDon't Share My Face: Privacy Preserving Inpainting for Visual Localization
Himmi, S., Ilter, O., Pailleau, F., Siegwart, R., Bescos, B., & Cadena, C. (2022) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros47612.2022.9982235 · Cited by 5
“After this first paper, the authors introduced uncertainty and a new geometric re-projection error into the model [13], [14].”
Spherical Panoramic Image-based Localization by Deep Learning
UMEDA, M., & Hisashi, D. A. T. E. (2018) · Transactions of the Society of Instrument and Control Engineers · doi:10.9746/sicetr.54.483 · Cited by 5
Bayesian Information Recovery from CNN for Probabilistic Inference
Kopitkov, D., & Indelman, V. (2018) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2018.8594506 · Cited by 5
“Network architecture parameters NF , NH and PD were sampled from ranges [2, 8], [100, 1000] and [0.”
BackgroundMethodInfluentialBoosting Image-Based Localization Via Randomly Geometric Data Augmentation
Wan, Y., Gao, W., Han, S., & Wu, Y. (2020) · Proceedings - International Conference on Image Processing · doi:10.1109/icip40778.2020.9190809 · Cited by 5
“Bayesian PoseNet[15] Learned β Posenet[7] Posenet Geo[7] LSTM pose[5] Hourglass Pose[4] MapNet +PGO[10] BranchNet [6] Relative Pose[9] RelocNet Fig.”
BackgroundParametric optimization of unmanned vehicle controller by PSO algorithm
Daryina, A. N., & Витальевич, П. И. (2021) · Procedia Computer Science · doi:10.1016/j.procs.2021.04.200 · Cited by 5
Enhancing the AR Experience with Machine Learning Services
Englert, M., Klomann, M., Weber, K., Grimm, P., & Jung, Y. (2019) · international conference on 3D web technology · doi:10.1145/3329714.3338134 · Cited by 5
“Recently, CNNs have also been used for estimating the camera pose using a single image for camera relocalization [Kendall and Cipolla 2015; Kendall et al. 2015; Wu et al. 2017].”
MethodPose-GNN : Camera Pose Estimation System Using Graph Neural NetworksPreprint
Elmogy, A., Dong, X., Lű, T., Westendorp, R., & Reddy, K. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2103.09435 · Cited by 9
“TABLE I THE MEDIAN ERROR IN POSITION (M)/ ORIENTATION (DEGREES) FOR THE 7 SCENES AND CAMBRIDGE DATASET FOR THE PROPOSED MODELS, COMPARED WITH SURFCNN [28], SURF-LSTM [43], POSENET [12], G-POSENET [44], POSENET-U [45], POSE-L [23], [46] AND G-POSENET [40], BRANCHNET [46] ,MOBILE-POSENET [47],VIDLOC [27] AND POSE-HOURGL…”
BackgroundNight-time indoor relocalization using depth image with Convolutional Neural Networks
Li, R., Liu, Q., Gui, J., Gu, D., & Hu, H. (2016) · International Conference on Automation and Computing (ICAC) · doi:10.1109/iconac.2016.7604929 · Cited by 4
“Recently there are also some researchers who began to use CNNs to solve pose regression and estimation problems [7] [8] without the need of using geometry methods.”
BackgroundA Deep Convolutional Neural Network Based 6-DOF Relocalization with Sensor Fusion System
Jo, H., Cho, H. M., Lee, S., & Kim, E. (2019) · The Journal of Korea Robotics Society · doi:10.7746/jkros.2019.14.2.087 · Cited by 4
Real-to-Synthetic Feature Transform for Illumination Invariant Camera Localization
Shoman, S., Mashita, T., Plopski, A., Ratsamee, P., & Uranishi, Y. (2020) · IEEE Computer Graphics and Applications · doi:10.1109/mcg.2020.3041617 · Cited by 4
Siddiq, M. I., Rizal, A., Erfianto, B., & Hadiyoso, S. (2023) · Applied Mechanics and Materials · doi:10.4028/p-9cplm1 · Cited by 4
Improving Learning-based Camera Pose Estimation for Image-based Augmented Reality Applications
Cai, E., Rossi, R. A., & Xiao, C. (2023) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3544549.3585756 · Cited by 4
“Regression-based methods use a single image as an input and output a camera pose directly via neural network [10, 11, 20].”
MethodScene-independent Localization by Learning Residual Coordinate Map with Cascaded Localizers
Wang, J., & Qi, Y. (2023) · Proceedings - International Symposium on Mixed and Augmented Reality, ISMAR · doi:10.1109/ismar59233.2023.00022 · Cited by 4
“Alternatively, learning features are used for localization by direct methods [9,23–25,39,54,57] or indirect pipelines [12,15,66].”
MethodCGAPoseNet+GCAN: A Geometric Clifford Algebra Network for Geometry-aware Camera Pose Regression
Pepe, A., Lasenby, J., & Buchholz, S. (2024) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv57701.2024.00646 · Cited by 5
“…and the geometric product, which, given two GA vectors u , v , is defined as in which · indicates the inner product and ∧ indicates the Grassman outer product. uv is a multivector, as it is a linear combination of objects with different grades, i.e. a scalar ( u · v ) and a bivector ( u ∧ v ).”
BackgroundMethodNavigating the Unknown: Uncertainty-Aware Compute-in-Memory Autonomy of Edge Robotics
Darabi, N., Shukla, P., Jayasuriya, D., Kumar, D., Stutts, A. C., & Trivedi, A. R. (2024) · Design, Automation, and Test in Europe · doi:10.23919/date58400.2024.10546628 · Cited by 4
Fan, Z., Dai, K., Guo, J., Jiang, Z., Gao, H., Xie, T., Li, R., & Wang, K. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3551843 · Cited by 4
“Technically, absolutely pose regression (APR) [30], [31], [32], relative pose regression (RPR) [4], [33], [34], and SCoRe [13], [15], [16], [17], [18], [35] compose the learning-based visual localization.”
Deep-mapnets : A residual network for 3D environment representation
Kuse, M., Jaiswal, S. P., & Shen, S. (2017) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2017.8296763 · Cited by 4
“[8] introduced a re-localization system based on a convolutional neural network to regress the camera pose.”
BackgroundThe Relocalization of SLAM Tracking Based on Spherical Cameras
Chang, Q., Liu, Q., Xin, Y., Huang, Y., Ren, F., & Cui, Y. (2021) · IEEE Access · doi:10.1109/access.2021.3130928 · Cited by 4
“And some recent researchers proposed CNNs-based approaches, they have shown great success in image-based localization, like PoseNet [9] and its variants [10]–[12].”
BackgroundFast and robust active camera relocalization in the wild for fine-grained change detection
Zhang, Q., Feng, W., Shi, Y., & Lin, D. (2022) · Neurocomputing · doi:10.1016/j.neucom.2022.04.102 · Cited by 4
Accurate visual localization with semantic masking and attention
Li, T., Zhan, Z., & Tan, G. (2022) · EURASIP Journal on Advances in Signal Processing · doi:10.1186/s13634-022-00875-2 · Cited by 4
“PoseNet [9] is the first end-toend network based on the GoogLeNet to regress the camera pose; follow-up approaches improve the accuracy with enhanced backbones [10, 11].”
MethodPal, S., Gupta, S., Das, N., & Ghosh, K. (2022) · Journal of Autonomous Vehicles and Systems · doi:10.1115/1.4055161 · Cited by 4
Ullah, M., Li, X., Hassan, M. A., Ullah, F., Muhammad, Y., Granelli, F., Vilčeková, L., & Sadad, T. (2022) · Sensors · doi:10.3390/s23010275 · Cited by 4
“For developing an efficient mobile or computer-based application, third-party libraries play a vital role in time minimization, providing an efficient interface, data management, algorithm development, and easy communication between multiple modules [26].”
BackgroundCNN-BASED INITIAL LOCALIZATION IMPROVED BY DATA AUGMENTATION
Mueller, M. S., Metzger, A., & Jutzi, B. (2018) · ISPRS annals of the photogrammetry, remote sensing and spatial information sciences · doi:10.5194/isprs-annals-iv-1-117-2018 · Cited by 4
“An enhancement is the Bayesian PoseNet (Kendall and Cipolla, 2016) which provides re-localization uncertainty by adding dropout layers after each convolutional layer and improves accuracy by averaging over multiple forward passes.”
MethodA Distributed Sensing Approach for Single Platform Image-Based Localization
Akal, O., Mukherjee, T., Barbu, A., Paquet, J., George, K., & Pasiliao, E. L. (2018) · International Conference on Machine Learning and Applications · doi:10.1109/icmla.2018.00103 · Cited by 4
“[2], [24] used Bayesian convolutional neural networks for both indoor and outdoor localization.”
MethodEuler Angles Based Loss Function for Camera Localization with Deep Learning
Fang, Q., Zhao, K., Tang, D., Zhou, Z., Zhou, Y., Hu, T., & Zhou, H. (2018) · IEEE International Conference on Cyber Technology in Automation, Control, and Intelligent Systems · doi:10.1109/cyber.2018.8688359 · Cited by 4
“The proposed deep learning models[7-9] regress the translation and the orientation(quaternion-based or the trigonometric function of the Euler angle) simultaneously, which have the inherent problem for the nondimensionalize of the orientation representation.”
MethodImage-similarity-based Convolutional Neural Network for Robot Visual Relocalization
Wang, L., Li, R., Sun, J., Seah, H. S., Quah, C. K., Zhao, L., & Tandianus, B. (2020) · Sensors and Materials · doi:10.18494/sam.2020.2549 · Cited by 4
Estimation of Linux Kernel Execution Path Uncertainty for Safety Software Test Coverage
Allende, I., Guire, N. M., Perez-Cerrolaza, J., Monsalve, L. G., Fernández, J. H., & Obermaisser, R. (2021) · Design, Automation, and Test in Europe · doi:10.23919/date51398.2021.9473951 · Cited by 4
Localising Faster: Efficient and precise lidar-based robot localisation in large-scale environments
Sun, L., Adolfsson, D., Magnusson, M., Andreasson, H., Posner, I., & Duckett, T. (2020) · White Rose Research Online (University of Leeds, The University of Sheffield, University of York) · doi:10.48550/arxiv.2003.01875 · Cited by 4
“Deep learning methods [8], [9], [10], [11], [12] are emerging in imagebased relocalisation as a pivotal precursor to directly estimate the 6-DOF pose based on a model learned for a specific environment with a deep regression neural network.”
BackgroundMethodFusion or not: Learning visual relocalization with matrix Fisher distribution
Fu, M., Li, S., Jin, Y., Zhang, W., Sychou, U., Skobtsov, V. Y., Sobolevskii, V., & Sokolov, B. (2024) · Neurocomputing · doi:10.1016/j.neucom.2024.129033 · Cited by 3
Beyond familiar landscapes: Exploring the limits of relative pose regressors in new environments
Idan, O., Shavit, Y., & Keller, Y. (2026) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2025.104629 · Cited by 3
Yao, D., Zhu, H., Ren, B., & Zhuang, X. (2024) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2024.108951 · Cited by 3
Predicting Out-of-View Feature Points for Model-Based Camera Pose Estimation
Moolan-Feroze, O., & Calway, A. D. (2018) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2018.8594297 · Cited by 3
“The authors expand on this in [3], where they explore the use of Bayesian Deep Learning to provide a measure of uncertainty alongside the pose estimate.”
BackgroundAn efficient image-based indoor positioning approach using ORB and LSH
Zhou, X., Meng, B., Dong, Y., Huang, X., & Zhang, K. (2021) · China Automation Congress (CAC) · doi:10.1109/cac53003.2021.9727606 · Cited by 3
A Deep Hierarchical Framework for Robot Global Localization
Wang, Z., Gao, F., Zhao, Y., Yin, Y., & Wang, L. (2022) · Journal of Intelligent & Robotic Systems · doi:10.1007/s10846-022-01728-8 · Cited by 3
“The authors further propose Bayesian PoseNet [18].”
CertainOdom: Uncertainty Weighted Multi-task Learning Model for LiDAR Odometry Estimation
Sun, L., Ding, G., Yoshiyasu, Y., & Kanehiro, F. (2022) · IEEE International Conference on Robotics and Biomimetics (ROBIO) · doi:10.1109/robio55434.2022.10011808 · Cited by 3
“The study of [18] uses the method of epistemic uncertainty estimation, which adds the dropout layer before each weight layer for both the training stage and inference stage.”
MethodPixel-Wise Prediction based Visual Odometry via Uncertainty Estimation
Chen, H., Liao, T.-H., Yang, H.-K., & Lee, C. (2023) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv56688.2023.00255 · Cited by 3
“implicitly captures the noises and enables the models to respond to the measured stochasticity inherent in the observations [11, 22, 40, 41, 42, 43, 44, 45].”
Monocular 6-DoF Pose Estimation for Non-cooperative Spacecrafts Using Riemannian Regression Network
Chu, S., Duan, Y., Schilling, K., & Wu, S. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-25056-9_13 · Cited by 4
Uncertainty Estimation for 3D Dense Prediction via Cross-Point Embeddings
Cai, K., Lu, C. X., & Huang, X. (2023) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2023.3256085 · Cited by 3
Robot Localization Using Situational Graphs (S-Graphs) and Building Architectural Plans
Shaheer, M., Bavle, H., Sánchez-López, J. L., & Voos, H. (2023) · Robotics · doi:10.3390/robotics12030065 · Cited by 3
“This is sometimes called the kidnapped robot prolem [3], re-localization [4], or place recognition [5].”
BackgroundConditional Variational Autoencoders for Probabilistic Pose Regression
Zangeneh, F., Bruns, L., Dekel, A., Pieropan, A., & Jensfelt, P. (2024) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros58592.2024.10802091 · Cited by 3
“Existing works on the task of visual relocalization commonly assess accuracy through median translation and orientation errors [7, 23, 31].”
Leveraging Spatial Attention and Edge Context for Optimized Feature Selection in Visual Localization
Istighfarin, N. F., & Jo, H. (2025) · International Journal of Control Automation and Systems · doi:10.1007/s12555-024-0487-y · Cited by 3
“Various methods have already been developed in APR including [5,26,27].”
SurfCNN: A descriptor enhanced convolutional neural network
Elmogy, A., Dong, X., Lű, T., Westendorp, R., & Reddy, K. (2019) · IEEE Pacific Rim Conference on Communications, Computers and Signal Processing (PACRIM) · doi:10.1109/pacrim47961.2019.8985026 · Cited by 3
SIR-Net: Scene-Independent End-to-End Trainable Visual Relocalizer
Nakashima, R., & Seki, A. (2019) · International Conference on 3D Vision (3DV) · doi:10.1109/3dv.2019.00059 · Cited by 3
“Note that the existing end-to-end trainable CNNs for relocalization [16, 14, 15, 44, 43, 25, 6, 7] take a scenedependent approach, thus these CNNs cannot be adapted to a new scene without re-training.”
BackgroundViPNet: An End-to-End 6D Visual Camera Pose Regression Network
Hu, H., Wang, A., Sons, M., & Lauer, M. (2020) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc45102.2020.9294630 · Cited by 3
“Later works extend the PoseNet to interpret relocalization uncertainty with Bayesian Neural Networks [6] and train the PoseNet with geometric loss function [7].”
BackgroundMethodCamera Pose Estimation using CNN
Wattanacheep, B., & Chitsobhuk, O. (2020) · International Conference on Control and Computer Vision (ICCCV) · doi:10.1145/3425577.3425593 · Cited by 3
“The final source (Outdoor: Cambridge dataset) [32] consists of 6 sub-datasets, GreatCourt, KingsCollege, OldHospital, ShopFacade, StMarysChurch and Street.”
BackgroundLARGER RECEPTIVE FIELD BASED RGB VISUAL RELOCALIZATION METHOD USING CONVOLUTIONAL NETWORK
Qin, J., Li, M., Li, D., Liao, X., Zhong, J., & Zhang, H. (2021) · The international archives of the photogrammetry, remote sensing and spatial information sciences/International archives of the photogrammetry, remote sensing and spatial information sciences · doi:10.5194/isprs-archives-xliii-b2-2021-47-2021 · Cited by 3
FFSCore-LSTM: An enhanced LSTM-based camera relocalization networks via front feature smoothing core
Wang, D., Liu, G., Prasad, B. D., Xiao, T., & Yang, Y. (2023) · Measurement · doi:10.1016/j.measurement.2023.112542 · Cited by 3
Towards accurate kidnap resolution through deep learning
Sommer, K., Kim, K.-H., Kim, Y., & Jo, S. (2017) · International Conference on Ubiquitous Robots and Ambient Intelligence · doi:10.1109/urai.2017.7992654 · Cited by 3
“PoseNet [14, 15] formulates the localization task as a six degree-of-freedom (DoF) regression problem.”
MethodElmogy, A., Dong, X., Lű, T., Westendorp, R., & Reddy, K. (2020) · IEEE Vehicular Technology Conference · doi:10.1109/vtc2020-fall49728.2020.9348762 · Cited by 3
“TABLE II THE MEDIAN ERROR IN POSITION (M)/ ORIENTATION (DEGREES) FOR THE 7 SCENES AND CAMBRIDGE DATASET FOR THE PROPOSED MODELS, COMPARED WITH SURFCNN [11], SURF-LSTM [23], POSENET [9], G-POSENET [26], POSENET-U [27], POSE-L [24], [28] AND G-POSENET [10], BRANCHNET [28] ,MOBILE-POSENET [29],VIDLOC [30] AND POSE-HOURGL…”
BackgroundSURF-LSTM: A Descriptor Enhanced Recurrent Neural Network For Indoor Localization
Elmogy, A., Dong, X., Lű, T., Westendorp, R., & Reddy, K. (2020) · IEEE Vehicular Technology Conference · doi:10.1109/vtc2020-fall49728.2020.9348748 · Cited by 3
“SURF-LSTM achieves an average error of 0.29 m using only 50 features which is slightly better than SurfCNN [20] that uses 300 features or comparable to PoseNet [11], PoseNet-U [27], Pose-LSTM [12], Mobile-PoseNet [29] and BranchNet [28] that use cropped images with CNN pretrained networks.”
BackgroundMethodInfluentialGraph-Based Thermal–Inertial SLAM With Probabilistic Neural NetworksPreprint
Saputra, M. R. U., Lu, C. X., de Gusmão, P. P. B., Wang, B., Markham, A., & Trigoni, N. (2021) · IEEE Transactions on robotics · doi:10.48550/arxiv.2104.07196 · Cited by 51
“This assumption has been used in [29], [56] as well.”
MethodTransCamP: Graph Transformer for 6-DoF Camera Pose EstimationPreprint
Li, X., & Ling, H. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2105.14065 · Cited by 3
“The proposed network is evaluated against recent state-of-the-art camera relocalization networks, including single image-based absolute camera pose regression network PoseNet and its variants [17–19, 37] among which,”
MethodOnline deep Bingham network for probabilistic orientation estimation
Li, W., Liu, J., Hao, W., Liu, H., Ren, D., Wang, Y., & Chen, L. (2023) · IET Computer Vision · doi:10.1049/cvi2.12188 · Cited by 2
DL-PoseNet: A Differential Lightweight Network for Pose Regression over SE(3)
Li, W., Liu, J., Wang, Y., Hao, W., Ren, D., & Chen, L. (2024) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra57147.2024.10611375 · Cited by 2
“…the accuracy of our lightweight model, we take a comparison with the state-of-the-art including the PoseNet and DensePoseNet [10], MapNet and MapNet++ [35], BPN [36], UBN and MBN-MB [2], AtLoc [23], GPoseNet [37], GNN-Pose [22] in terms of final evaluation results on two common datasets.”
MethodPose Estimation of Mobile Robot Using Image and Point-Cloud Data
An, S. W., & Park, H. S. (2024) · Journal of Electrical Engineering and Technology · doi:10.1007/s42835-024-02030-3 · Cited by 2
“To compare with previous studies, deep-learning-based pose-estimation models, such as PoseNet [9], PoseNet-Bayesian [10], and PoseLSTM [11] preprocessing, duplicate images were removed, reducing the total number from 145,000 to 14,314.”
InfluentialAlam, M. S., Mohamed, F., & Hossain, A. B. (2025) · Computer Science Journal of Moldova · doi:10.56415/csjm.v33.04 · Cited by 2
A method for absolute pose regression based on cascaded attention modules
Song, X., Tang, J., Yang, K., Guo, W., Lu, X., & Hei, X. (2025) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2025.104440 · Cited by 2
Logistics Planning of Autonomous Air Cargo Vehicles with Deep Learning Methods: A Literature Review
Gör, M. S., & Çelik, C. (2025) · Applied Sciences · doi:10.3390/app151910709 · Cited by 3
Geometrical Features based Visual Relocalization for Indoor Service Robot
Wang, J., Dai, D., Wang, P., Zhao, H., Xu, M., & Chen, Z. (2020) · International Conference on Communication, Image and Signal Processing (CCISP) · doi:10.1109/ccisp51026.2020.9273485 · Cited by 2
Scale-invariant localization using quasi-semantic object landmarks
Holliday, A. G. C., & Dudek, G. (2021) · Autonomous Robots · doi:10.1007/s10514-021-09973-w · Cited by 2
“Kendall and Cipolla (2016) train a CNN to predict camera pose from an image of an outdoor environment.”
Background
Metadata
- Authors
- Kendall, A., & Cipolla, R.
- Affiliations
- University of Cambridge (GB)
- ORCID
- Cipolla, Roberto
- Venue
- 2016 IEEE International Conference on Robotics and Automation (ICRA)
- Publisher
- IEEE
- Published
- 2016-05
- Pages
- 4762-4769
- ISSN
- 1050-4729 (ISSN-L)
- OpenAlex
- W2279895976
- Cited by
- 548
- References
- 30
- Citations by year
- 2026: 20 · 2025: 32 · 2024: 44 · 2023: 57 · 2022: 52 · 2021: 68 · 2020: 90 · 2019: 82 · 2018: 57 · 2017: 39 · 2016: 6
- FWCI
- 107.26
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- Proceedings - IEEE International Conference on Robotics and Automation/Proceedings (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- Cambridge University Engineering Department Publications Database (submittedVersion)
- Apollo (University of Cambridge) (submittedVersion)
- arXiv (Cornell University)
- Apollo
- 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 Image and Video Retrieval Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Vision and Imaging
- Keywords
- camera relocalization, Bayesian convolutional neural network, uncertainty estimation, 6D pose estimation, real-time localization, out-of-distribution detection, indoor and outdoor scenes
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
