3D Pose Regression Using Convolutional Neural Networks
Open accessThe DOI matches its Crossref record; 21 of the 24 references it cites were found in a registry or an index.
Abstract
3D pose estimation is a key component of many important computer vision tasks such as autonomous navigation and 3D scene understanding. Most state-of-the-art approaches to 3D pose estimation solve this problem as a pose-classification problem in which the pose space is discretized into bins and a CNN classifier is used to predict a pose bin. We argue that the 3D pose space is continuous and propose to solve the pose estimation problem in a CNN regression framework with a suitable representation, data augmentation and loss function that captures the geometry of the pose space. Experiments on PASCAL3D+ show that the proposed 3D pose regression approach achieves competitive performance compared to the state-of-the-art.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
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 cited source doesn’t support the claim as written.
In the citing paper
“Geodesic loss, L geo , which is the square of the geodesic distance in SE(3), has been studied for registration problems due to its convexity with respect to pose transformations [32] [33].”
From A Fully Differentiable Framework for 2D/3D Registration and the Projective Spatial Transformers (2024) · cites this paper as [33]
Checked by a reasoning model — may be wrong.
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“Quaternions, for example, are non-injective. While one can map quaternions to lie only on one hemisphere as in [38], L 2 distances to quaternions near the equator will still not express the geodesic distance.”
From Deep Learning for Spacecraft Pose Estimation from Photorealistic Rendering (2020) · cites this paper as [38]
The check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“It is worth noting that we have experimented using the adaptive weighting based on Laplace likelihood in [38] but achieved poor results.”
From Deep Learning for Spacecraft Pose Estimation from Photorealistic Rendering (2020) · cites this paper as [38]
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
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
Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Ren, S., He, K., Girshick, R., & Sun, J. (2017) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2016.2577031 · Cited by 73,472
ImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333
Return of the Devil in the Details: Delving Deep into Convolutional Nets
Chatfield, K., Simonyan, K., Vedaldi, A., & Zisserman, A. (2014) · Proceedings of the British Machine Vision Conference 2014 · doi:10.5244/c.28.6 · Cited by 3,335
XSEDE: Accelerating Scientific Discovery
Towns, J., Cockerill, T., Dahan, M., Foster, I., Gaither, K., Grimshaw, A., Hazlewood, V., Lathrop, S., Lifka, D., Peterson, G. D., et al. (2014) · Computing in Science & Engineering · doi:10.1109/mcse.2014.80 · Cited by 3,109
PoseNet: 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
Geometric Loss Functions for Camera Pose Regression with Deep Learning
Kendall, A., & Cipolla, R. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.694 · Cited by 803 · cited 1×
“Quaternions have also been used to represent 3D pose for camera localization in [11, 10, 9], but these works ignore the unit-norm constraint for computational ease and use a mean-squared or reprojection loss, whereas we incorporate the constraint into the network and use a geodesic loss.”
MethodBeyond PASCAL: A benchmark for 3D object detection in the wild
Xiang, Y., Mottaghi, R., & Savarese, S. (2014) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv.2014.6836101 · Cited by 727
Render for CNN: Viewpoint Estimation in Images Using CNNs Trained with Rendered 3D Model Views
Su, H., Qi, C. R., Li, Y., & Guibas, L. J. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.308 · Cited by 680
Modelling uncertainty in deep learning for camera relocalization
Kendall, A., & Cipolla, R. (2016) · 2016 IEEE International Conference on Robotics and Automation (ICRA) · doi:10.1109/icra.2016.7487679 · Cited by 548 · cited 1×
“Quaternions have also been used to represent 3D pose for camera localization in [11, 10, 9], but these works ignore the unit-norm constraint for computational ease and use a mean-squared or reprojection loss, whereas we incorporate the constraint into the network and use a geodesic loss.”
Method
Research works printed but not linked: 4
- ma, 2003, An Invitation to 3-D Vision from Images to Geometric Models
- simonyan, 2014, Very Deep Convolutional Networks for Large-scale Image Recognition
- 2015, TensorFlow Large-Scale Machine Learning on Heterogeneous Systems
- 2015, Keras
Cited by
Real-Time Seamless Single Shot 6D Object Pose Prediction
Tekin, B., Sinha, S. N., & Fua, P. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00038 · Cited by 900
Implicit 3D Orientation Learning for 6D Object Detection from RGB Images
Sundermeyer, M., Marton, Z.-C., Durner, M., Brucker, M., & Triebel, R. (2018) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-01231-1_43 · Cited by 564
“In practice, direct regression approaches for full 3D object orientation estimation have not been very successful [23].”
BackgroundApplication of deep learning algorithms in geotechnical engineering: a short critical review
Zhang, W., Li, H., Li, Y., Liu, H., Chen, Y., & Ding, X. (2021) · Artificial Intelligence Review · doi:10.1007/s10462-021-09967-1 · Cited by 538
“…(Shi et al. 2018),speech recognition (Zhang et al. 2017c), human behavior recognition (Han et al. 2018; Sargano et al. 2017), 3D pose estimation (Mahendran et al. 2017), plant disease identification (Liu et al. 2018; Lu et al. 2017), medical image analysis and applications (Havaei et al. 2017;…”
BackgroundLi, Z., Wang, G., & Ji, X. (2019) · 2019 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00777 · Cited by 485
“Direct approaches were initially proposed in some works on 6-DoF camera pose estimation and viewpoint estimation [22, 11, 10, 16, 23, 14].”
Satellite Pose Estimation Challenge: Dataset, Competition Design, and Results
Kisantal, M., Sharma, S., Park, T. H., Izzo, D., Martens, M., & D'Amico, S. (2020) · IEEE Transactions on Aerospace and Electronic Systems · doi:10.1109/taes.2020.2989063 · Cited by 304
Spacecraft pose estimation dataset (speed)
Sharma, S., Park, T. H., & D'Amico, S. (2026) · Stanford Digital Repository · doi:10.25740/dz692fn7184 · Cited by 304
RelocNet: Continuous Metric Learning Relocalisation Using Neural Nets
Balntas, V., Li, S., & Prisacariu, V. (2018) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-01264-9_46 · Cited by 223
Deep Learning for Spacecraft Pose Estimation from Photorealistic Rendering
Proenca, P. F., & Gao, Y. (2020) · 2020 IEEE International Conference on Robotics and Automation (ICRA) · doi:10.1109/icra40945.2020.9197244 · Cited by 229
“[21] also regress the angle-axis vector, but they minimize directly the geodesic loss.”
BackgroundMethodNeural Network-Based Pose Estimation for Noncooperative Spacecraft Rendezvous
Sharma, S., & D'Amico, S. (2020) · IEEE Transactions on Aerospace and Electronic Systems · doi:10.1109/taes.2020.2999148 · Cited by 196
“The deep learning-based methods either discretize the pose space and solve the resulting classification problem [42]–[45] or directly regress the relative pose from the input image [46]–[48].”
MethodAugmented Autoencoders: Implicit 3D Orientation Learning for 6D Object Detection
Sundermeyer, M., Márton, Z.-C., Durner, M., & Triebel, R. (2019) · International Journal of Computer Vision · doi:10.1007/s11263-019-01243-8 · Cited by 183
“(2015), Mitash et al. (2017) and Richter et al. (2016) have shown that photo-realistic renderings of object views and backgrounds can in some cases benefit the generalization performance for tasks like object detection and viewpoint estimation.”
BackgroundMethodInfluentialThree-D Safari: Learning to Estimate Zebra Pose, Shape, and Texture From Images “In the Wild”
Zuffi, S., Kanazawa, A., Berger-Wolf, T., & Black, M. (2019) · 2019 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00546 · Cited by 139
Real-Time Deep Pose Estimation With Geodesic Loss for Image-to-Template Rigid Registration
Mohseni Salehi, S. S., Khan, S., Erdogmus, D., & Gholipour, A. (2019) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2018.2866442 · Cited by 115
“[21] have recently modeled the 3D camera/object pose estimation as a regression problem.”
MethodResultA Vector-based Representation to Enhance Head Pose Estimation
Cao, Z., Chu, Z., Liu, D., & Chen, Y. (2021) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv48630.2021.00123 · Cited by 154
“The other type of methods such as [30, 16, 7] use network to predict orientation directly.”
MethodReal-time Deep Pose Estimation with Geodesic Loss for Image-to-Template Rigid Registration
Salehi, S. S. M., Khan, S., Erdogmus, D., & Gholipour, A. (2018) · arXiv · doi:10.48550/arxiv.1803.05982 · Cited by 118
“[20] have recently modeled the 3D camera/object pose estimation as a regression problem.”
MethodMulti-Path Learning for Object Pose Estimation Across Domains
Sundermeyer, M., Durner, M., Puang, E. Y., Márton, Z.-C., Vaškevičius, N., Arras, K. O., & Triebel, R. (2020) · 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/cvpr42600.2020.01393 · Cited by 133
Image-based velocity estimation of rock using Convolutional Neural Networks
Karimpouli, S., & Tahmasebi, P. (2019) · Neural Networks · doi:10.1016/j.neunet.2018.12.006 · Cited by 109
Bukschat, Y., & Vetter, M. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2011.04307 · Cited by 66
“[24] found that it also performed slightly better in their experiments.”
Background6D Object Pose Regression via Supervised Learning on Point Clouds
Gao, G., Lauri, M., Wang, Y., Hu, X., Zhang, J., & Frintrop, S. (2020) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra40945.2020.9197461 · Cited by 114
“Axis-angle representation is also used in [28].”
Phisannupawong, T., Kamsing, P., Torteeka, P., Channumsin, S., Sawangwit, U., Hematulin, W., Jarawan, T., Somjit, T., Yooyen, S., Delahaye, D., et al. (2020) · Aerospace · doi:10.3390/aerospace7090126 · Cited by 83
“In many studies [14,18,20], the orientation loss was formulated as the geodesic loss in Equation (2).”
BackgroundMethodA Geometric Approach to Obtain a Bird's Eye View From an Image
Abbas, S. A., & Zisserman, A. (2019) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw.2019.00504 · Cited by 72
Towards Robust Learning-Based Pose Estimation of Noncooperative Spacecraft
Park, T. H., Sharma, S., & D'Amico, S. (2019) · arXiv · doi:10.48550/arxiv.1909.00392 · Cited by 65
6D object position estimation from 2D images: a literature review
Marullo, G., Tanzi, L., Piazzolla, P., & Vezzetti, E. (2023) · Multimedia Tools and Applications · doi:10.1007/s11042-022-14213-z · Cited by 64
“The method proposed in [30] focused only on rotation between the object and the camera using a modified version of the VGG-M network.”
MethodOptimizing Through Learned Errors for Accurate Sports Field Registration
Jiang, W., Higuera, J. C. G., Angles, B., Sun, W., Javan, M., & Yi, K. M. (2020) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv45572.2020.9093581 · Cited by 73
“Deep networks have also been proposed to directly regress the 6 DoF pose of cameras [51, 43, 25].”
BackgroundThree dimensional objects recognition & pattern recognition technique; related challenges: A review
Rani, S., Lakhwani, K., & Kumar, S. (2022) · Multimedia Tools and Applications · doi:10.1007/s11042-022-12412-2 · Cited by 98
Liebmann, F., von Atzigen, M., Stütz, D., Wolf, J., Zingg, L., Suter, D., Cavalcanti, N., Leoty, L., Esfandiari, H., Snedeker, J. G., et al. (2023) · Medical Image Analysis · doi:10.1016/j.media.2023.103027 · Cited by 69
“As in Tulsiani and Malik [2015] and Mahendran et al. [2017], the orientation prediction was evaluated with the median geodesic angle error (MGAE), which equals the median loss defined in Equation (1) over an entire fold, expressed in degrees.”
MethodAn Analysis of SVD for Deep Rotation EstimationPreprint
Levinson, J., Esteves, C., Chen, K., Snavely, N., Kanazawa, A., Rostamizadeh, A., & Makadia, A. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2006.14616 · Cited by 138
“The task of predicting 3D rotations is common to estimating object pose [47, 24, 29, 38, 43, 21, 39], relative camera pose [27, 32, 5], ego-motion and depth from video [48, 26], and human pose [49, 19].”
BackgroundSpherical Regression: Learning Viewpoints, Surface Normals and 3D Rotations on N-Spheres
Liao, S., Gavves, E., & Snoek, C. G. M. (2019) · 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/cvpr.2019.00999 · Cited by 80
“[22] consider quaternion for camera re-localization and viewpoint estimation.”
InfluentialConvolutional Neural Network-Based Visual Servoing for Eye-to-Hand Manipulator
Tokuda, F., Arai, S., & Kosuge, K. (2021) · IEEE Access · doi:10.1109/access.2021.3091737 · Cited by 55
FADN: Fully Connected Attitude Detection Network Based on Industrial Video
Yang, J., Xi, M., Jiang, B., Man, J., Meng, Q., & Li, B. (2020) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2020.2984370 · Cited by 51
“Atthesame time, the attitude angle estimation could also be divided into the visual correspondence based algorithms [5] and the deep learning based methods [6].”
BackgroundDeep Bingham Networks: Dealing with Uncertainty and Ambiguity in Pose Estimation
Deng, H., Bui, M., Navab, N., Guibas, L. J., Ilić, S., & Birdal, T. (2022) · International Journal of Computer Vision · doi:10.1007/s11263-022-01612-w · Cited by 76
“A metric often used for measuring distance on the spherical manifold, cosine loss respects the vectorial nature of quaternions [65]: Lcos = 1− |q · q̂|.”
Multi-person 3D Pose Estimation in Crowded Scenes Based on Multi-view Geometry
Chen, H., Guo, P., Li, P., Lee, G. H., & Chirikjian, G. S. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-58580-8_32 · Cited by 67
Deep Predictive Motion Tracking in Magnetic Resonance Imaging: Application to Fetal Imaging
Singh, A., Salehi, S. S. M., & Gholipour, A. (2020) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2020.2998600 · Cited by 58
“While the majority of pose estimation techniques have been designed as classification methods, the problem has been recently modeled and solved by regression deep neural networks [31].”
BackgroundLei, H., Liu, S., Elazab, A., Gong, X., & Lei, B. (2021) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2020.3027566 · Cited by 46
“In this work, we extensively test different models and eventually select the VGG16 [7] as the backbone network, initialized with ImageNet pre-trained weights, since it has proven remarkable performance in image classification tasks [52]–[54].”
MethodGeneralizing Spatial Transformers to Projective Geometry with Applications to 2D/3D Registration
Gao, C., Liu, X., Gu, W., Killeen, B., Armand, M., Taylor, R., & Unberath, M. (2020) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-59716-0_32 · Cited by 46
Single Shot 6D Object Pose Estimation
Kleeberger, K., & Huber, M. F. (2020) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra40945.2020.9197207 · Cited by 59
“For classical approaches and CNN-based methods relying on region proposals [32], [1], [6], [7], [20] or a prior segmentation step [8], [13], [14], [15] for pose estimation, the runtime increases linearly with the number of objects present in the scene, which is not optimal for very cluttered scenes.”
BackgroundMethodWide-Baseline Relative Camera Pose Estimation with Directional Learning
Chen, K., Snavely, N., & Makadia, A. (2021) · 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/cvpr46437.2021.00327 · Cited by 79
“Unfortunately, evidence suggests designing regression models for pose estimation is challenging, and in fact finding a parameterization of the motion groups effective in deep learning models is still an active research topic [33, 66, 44].”
BackgroundDeep Learning for Spacecraft Pose Estimation from Photorealistic Rendering
Proenca, P. F., & Gao, Y. (2019) · arXiv · Cited by 251
“[21] also regress the angle-axis vector, but they minimize directly the geodesic loss.”
BackgroundMethodDrone Detection and Pose Estimation Using Relational Graph Networks
Jin, R., Jiang, J., Qi, Y., Lin, D., & Song, T. (2019) · Sensors · doi:10.3390/s19061479 · Cited by 43
“[20] tried to directly regress full 3D object orientation by natural non-linearity output layer and an appropriate geodesic loss, and they used fine labeled 3D object with pose PASCAL3D+ [27].”
MethodA Fully Differentiable Framework for 2D/3D Registration and the Projective Spatial Transformers
Gao, C., Feng, A., Liu, X., Taylor, R. H., Armand, M., & Unberath, M. (2024) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2023.3299588 · Cited by 55
“Geodesic loss, L geo , which is the square of the geodesic distance in SE ( 3 ) , has been studied for registration problems due to its convexity with respect to pose transformations [32], [33].”
BackgroundAn enhanced marker pattern that achieves improved accuracy in surgical tool tracking
Cartucho, J., Wang, C., Huang, B., Elson, D. S., Darzi, A. W., & Giannarou, S. (2021) · Computer Methods in Biomechanics and Biomedical Engineering Imaging & Visualization · doi:10.1080/21681163.2021.1997647 · Cited by 38
“Vision-based tracking methods include feature-based methods, and end-to-end tool pose estimation using deep learning (Kendall et al. 2015; Mahendran et al. 2017).”
MethodSelf-Supervised Viewpoint Learning From Image Collections
Mustikovela, S. K., Jampani, V., Mello, S. D., Liu, S., Iqbal, U., Rother, C., & Kautz, J. (2020) · 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/cvpr42600.2020.00403 · Cited by 43
Li, F., Wang, Y., Lü, H., Li, P., Li, Q., & Liu, F. (2020) · Journal of Physics G Nuclear and Particle Physics · doi:10.1088/1361-6471/abb1f9 · Cited by 44
Spin Detection in Robotic Table Tennis
Tebbe, J., Klamt, L., Gao, Y., & Zell, A. (2020) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra40945.2020.9196536 · Cited by 34
“Related work on pose detection with neural networks [14], [15] favours the residual network architecture from He et al.”
BackgroundMethodCao, Z., Liu, D., Wang, Q., & Chen, Y. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19775-8_43 · Cited by 46
Real-Time, Flight-Ready, Non-Cooperative Spacecraft Pose Estimation Using Monocular Imagery
Black, K., Shankar, S., Fonseka, D., Deutsch, J., Dhir, A., & Akella, M. R. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2101.09553 · Cited by 20
“Rather than modeling the problem as a classification task, Mahendran et al. (2017) 18 show that directly regressing a quaternion representation of pose is a viable method where minimal, if any, post-processing is required.”
Deep neural networks for human pose estimation from a very low resolution depth image
Szczuko, P. (2019) · Multimedia Tools and Applications · doi:10.1007/s11042-019-7433-7 · Cited by 35
“VGG and Alexnet were trained to estimate rigid object orientation in a 2D images, first applying classification network, and then dedicated pose network for each object class, inferring 3 orientation coordinates [17].”
Method3D Pose Estimation for Fine-Grained Object Categories
Wang, Y., Tan, X., Yang, Y., Liu, X., Ding, E., Zhou, F., & Davis, L. S. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-11009-3_38 · Cited by 33
Vision-Based Sensing for Electrically-Driven Soft Actuators
Zhang, A., Truby, R. L., Chin, L., Li, S., & Rus, D. L. (2022) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2022.3201604 · Cited by 34
A Mixed Classification-Regression Framework for 3D Pose Estimation from 2D ImagesPreprint
Mahendran, S., Ali, H., & Vidal, R. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1805.03225 · Cited by 46
“However, as recommended by [13], it makes more sense to minimize a geodesic loss on the space of rotation matrices instead of the Euclidean loss as it better captures the geometry of the problem.”
BackgroundMethodInfluentialSelf-citationProbabilistic orientation estimation with matrix Fisher distributionsPreprint
Mohlin, D., Bianchi, G., & Sullivan, J. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2006.09740 · Cited by 60
“Commonly, quaternions [28] or Euler angles [14] representation are used.”
MethodObject Pose Estimation from Monocular Image Using Multi-view Keypoint Correspondence
Kundu, J. N., Rahul, M., Ganeshan, A., & Babu, R. V. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-11015-4_23 · Cited by 31
Learning Local RGB-to-CAD Correspondences for Object Pose Estimation
Georgakis, G., Karanam, S., Wu, Z., & Košecká, J. (2019) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00906 · Cited by 34
“Furthermore, even though several methods have shown promising results on 3D object pose estimation with real RGB images, they either require accurate 3D annotations [23, 26, 44, 30, 16] or 3D object models with realistic textures [5, 37, 6, 17] in the training stage.”
BackgroundMethodXiao, Y., Du, Y., & Marlet, R. (2021) · International Conference on 3D Vision (3DV) · doi:10.1109/3dv53792.2021.00018 · Cited by 36
Shi, W., Xu, H., Sun, C., Sun, J., Li, Y., Xu, X., Zheng, T., Zhang, Y., Wang, G., & Wu, D. (2022) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2022.3208277 · Cited by 30
“, descriptor matching [27], [28] and absolute pose regression [29], [30], can be transferred to the problem of fetal motion estimation where the motion can be characterized by the continuous alternation of poses in the 3D canonical (atlas) space.”
BackgroundBenchmarking Domain Randomisation for Visual Sim-to-Real Transfer
Alghonaim, R., & Johns, E. (2021) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48506.2021.9561134 · Cited by 33
“For the orientation, however, we use Equation 1 [37] to find the angle required to rotate from one quaternion (the label, qi) to the other (the prediction, q̂i).”
Regression-Based Three-Dimensional Pose Estimation for Texture-Less Objects
Liu, Y., Zhou, L., Zong, H., Gong, X., Wu, Q., Liang, Q., & Wang, J. (2019) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2019.2913321 · Cited by 26
“Unlike [38], [42], they focus the feature representation in the regression task and use pose information to guide the feature extraction in the network to obtain an efficient regression model.”
Method3D Object Pose Estimation Using Multi-Objective Quaternion Learning
Papaioannidis, C., & Pitas, I. (2019) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2019.2929600 · Cited by 27
“Also, a pre-trained VGG-M [44] network on ImageNet was used as a base network, coupled with a trained pose estimation network in [45] to perform 3D pose regression.”
MethodDomain-Translated 3D Object Pose Estimation
Papaioannidis, C., Mygdalis, V., & Pitas, I. (2020) · IEEE Transactions on Image Processing · doi:10.1109/tip.2020.3025447 · Cited by 27
“CNNs [13], [30] or a separate 3D pose estimation network for each object of interest [15].”
BackgroundSU-Net: pose estimation network for non-cooperative spacecraft on-orbit
Gao, H., Li, Z., Wang, N., Yang, J., & Dang, D. (2023) · Scientific Reports · doi:10.1038/s41598-023-38974-1 · Cited by 28
Contact2Grasp: 3D Grasp Synthesis via Hand-Object Contact Constraint
Li, H., Lin, X., Zhou, Y. M., Li, X., Huo, Y., Chen, J., & Ye, Q. (2023) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2023/117 · Cited by 42
“We use the L1 loss for the translation parameter and the geodesic loss [ Mahendran et al. , 2017 ] for the pose parameter, named as L t and L p respectively.”
MethodLearning Deep Network for Detecting 3D Object Keypoints and 6D Poses
Zhao, W., Zhang, S., Guan, Z., Zhao, W., Peng, J., & Jianping Fan, B. (2020) · 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/cvpr42600.2020.01414 · Cited by 32
“We use a geodesic distance [21] as the lose function:”
MethodJo, H., & Kim, E. (2020) · IEEE Access · doi:10.1109/access.2020.2988464 · Cited by 22
“The position distance is the Euclidean norm of the position vector’s difference, and the distance between the two orientations represented by the rotation matrix uses the geodesic distance used in [43].”
MethodVisual Collaboration Leader-Follower UAV-Formation for Indoor Exploration
Evangeliou, N., Chaikalis, D., Τσουκαλάς, Α., & Tzes, A. P. (2022) · Frontiers in Robotics and AI · doi:10.3389/frobt.2021.777535 · Cited by 25
“More recently, deep learning techniques for relative pose estimation have been introduced for accurate results (Mahendran et al., 2017; Radwan et al., 2018; Patel et al., 2021).”
BackgroundEnhancing object pose estimation for RGB images in cluttered scenes
Al-Selwi, M., Huang, N., Yin, G., Chao, Y., Li, Q., & Li, J. (2025) · Scientific Reports · doi:10.1038/s41598-025-90482-6 · Cited by 22
“Some works 51 proved that axis-angle representation for rotation achieved significant performance 24 .”
Object Pose Estimation using Mid-level Visual Representations
Nejatishahidin, N., Fayyazsanavi, P., & Košecká, J. (2022) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros47612.2022.9981452 · Cited by 20
Pérez-Villar, J. I. B., García‐Martín, Á., & Bescós, J. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-25056-9_3 · Cited by 36
SGRec3D: Self-Supervised 3D Scene Graph Learning via Object-Level Scene Reconstruction
Koch, S., Hermosilla, P., Vaškevičius, N., Colosi, M., & Ropinski, T. (2024) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv57701.2024.00337 · Cited by 48
Occlusion Resistant Object Rotation Regression from Point Cloud Segments
Gao, G., Lauri, M., Zhang, J., & Frintrop, S. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-11009-3_44 · Cited by 22
A Geometric Approach to Obtain a Bird's Eye View from an ImagePreprint
Abbas, A. N., & Zisserman, A. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1905.02231 · Cited by 21
Pose-guided Auto-Encoder and Feature-Based Refinement for 6-DoF Object Pose Regression
Li, Z., & Ji, X. (2020) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra40945.2020.9196953 · Cited by 23
“[6], [7], [8], [9], [10] directly regressed the continuous pose value from image, but only achieved limited performance.”
BackgroundUpper Body Pose Estimation Using Deep Learning for a Virtual Reality Avatar
Anvari, T., Park, K., & Kim, G. (2023) · Applied Sciences · doi:10.3390/app13042460 · Cited by 22
Deep-Learning-Based Direct Attitude Estimation for Uncooperative Known Space Objects
Qiao, S., Zhang, H., Xie, F., & Jiang, Z. (2023) · IEEE Transactions on Aerospace and Electronic Systems · doi:10.1109/taes.2023.3325801 · Cited by 19
“The rotation matrix error defines a geodesic distance between two rotation matrices [65].”
BackgroundUncertainty-Aware Self-Supervised Learning of Spatial Perception Tasks
Nava, M., Paolillo, A., Guzzi, J., Gambardella, L. M., & Giusti, A. (2021) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2021.3095269 · Cited by 22
“In (5), ∆ quat ( · , · ) denotes the quaternionic distance [23, eq.”
BackgroundL6DNet: Light 6 DoF Network for Robust and Precise Object Pose Estimation With Small Datasets
Gonzalez, M., Kacete, A., Murienne, A., & Marchand, É. (2021) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2021.3062605 · Cited by 23
“In some recent works [19], [21], [22] the choice of representation for rotations has been studied as it shows to have an impact on the accuracy of the pose estimation [22].”
BackgroundDeep learning-based plane pose regression in obstetric ultrasound
Di Vece, C., Dromey, B., Vasconcelos, F., David, A. L., Peebles, D., & Stoyanov, D. (2022) · International Journal of Computer Assisted Radiology and Surgery · doi:10.1007/s11548-022-02609-z · Cited by 17
Robust Six Degrees of Freedom Estimation for IIoT Based on Multibranch Network
Yang, J., Xi, M., Jiang, B., & Song, H. (2020) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2020.2982703 · Cited by 17
“Recent years have witnessed a flourishing prosperity of the computer visual perception technology, and numerous deep learning-based approaches are gradually emerging [8], [9].”
BackgroundUltra-low Power Deep Learning-based Monocular Relative Localization Onboard Nano-quadrotors
Bonato, S., Lambertenghi, S. C., Cereda, E., Giusti, A., & Palossi, D. (2023) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48891.2023.10161127 · Cited by 16
“of rigid transformations in a machine learning context [16].”
BackgroundKVNet: An iterative 3D keypoints voting network for real-time 6-DoF object pose estimation
Wang, F., Zhang, X., Chen, T., Shen, Z., Liu, S., & He, Z. (2023) · Neurocomputing · doi:10.1016/j.neucom.2023.01.036 · Cited by 17
Canonical Voting: Towards Robust Oriented Bounding Box Detection in 3D Scenes
You, Y., Ye, Z., Lou, Y., Li, C., Li, Y., Ma, L., Wang, W., & Lu, C. (2022) · 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/cvpr52688.2022.00126 · Cited by 14
“ate instances. DenseFusion [21] predicts a rotation and confidence score densely for each pixel. At inference time, rotations with the highest confi- dence are used as the output. Both Mahendran et al. [12] and Gao et al. [4] propose to directly regress rotation from 2D images and 3D point clouds, respectively. 3. Metho…”
MethodInfluentialAlbanis, G., Zioulis, N., Dimou, A., Zarpalas, D., & Daras, P. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-66096-3_44 · Cited by 21
Vision-based attitude estimation for spacecraft docking operation through deep learning algorithm
Phisannupawong, T., Kamsing, P., Tortceka, P., & Yooyen, S. (2020) · International Conference on Advanced Communication Technology · doi:10.23919/icact48636.2020.9061445 · Cited by 16
“[11], [12] pose estimation frameworks are implemented on the Pascal dataset, which contain many kinds of objects for training in pose estimation field.”
BackgroundMethodPose Estimation for Non-Cooperative Rendezvous Using Neural Networks
Sharma, S., & D'Amico, S. (2019) · arXiv · Cited by 40
HCR-Net: A Hybrid of Classification and Regression Network for Object Pose Estimation
Wang, Z., Li, W., Kao, Y., Zou, D., Wang, Q., Ahn, M., & Hong, S. (2018) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2018/141 · Cited by 13
“Recent CNN based pose estimation methods formulate this problem by either multi-class classification [Su et al. , 2015; Tulsiani and Malik, 2015] or regression [Mahendran et al. , 2017; Fenzi et al. , 2013].”
MethodVision-state Fusion: Improving Deep Neural Networks for Autonomous Robotics
Cereda, E., Bonato, S., Nava, M., Giusti, A., & Palossi, D. (2024) · Journal of Intelligent & Robotic Systems · doi:10.1007/s10846-024-02091-6 · Cited by 12
“We repeat these tests for three subjects, leading to 30 trials for each experiment – 60 flight tests in total.”
Reconstructing the 3D digital core with a fully convolutional neural network
Li, Q., Chen, Z., He, J., Hao, S., Wang, R., Yang, H.-T., & Sun, H. (2020) · Applied Geophysics · doi:10.1007/s11770-020-0822-x · Cited by 12
Zhu, Y., Gao, R., Huang, S., Zhu, S. C., & Wu, Y. N. (2021) · 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/cvpr46437.2021.00983 · Cited by 12
“[36, 24] use axis-angle representation, which represents 3D orientation by the direction of axis of rotation as well as the magnitude of the rotation.”
MethodEfficient 6-DoF Tracking of Handheld Objects from an Egocentric Viewpoint
Pandey, R. K., Pidlypenskyi, P., Yang, S., & Kaeser‐Chen, C. (2018) · Lecture notes in computer science · doi:10.1007/978-3-030-01216-8_26 · Cited by 14
Novel Object Viewpoint Estimation Through Reconstruction Alignment
Banani, M. E., Corso, J. J., & Fouhey, D. F. (2020) · 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/cvpr42600.2020.00318 · Cited by 15
Matching RGB Images to CAD Models for Object Pose Estimation.Preprint
Georgakis, G., Karanam, S., Wu, Z., & Košecká, J. (2018) · arXiv (Cornell University) · Cited by 13
A Passenger Detection and Action Recognition System for Public Transport Vehicles
Lin, H., Kao, S.-F., & Wang, C. (2024) · Journal of Intelligent & Robotic Systems · doi:10.1007/s10846-024-02194-0 · Cited by 12
“Although the CNN-based algorithms have improved the recognition accuracy significantly [17], the computational cost and real-time processing required still remain as unsolved problems.”
Arza-García, M., Gonçalves, J. A., Pinto, V. F., & Bastos, G. (2024) · Remote Sensing · doi:10.3390/rs16020331 · Cited by 10
Optimal STEM Convergence Angle Selection Using a Convolutional Neural Network and the Strehl Ratio
Schnitzer, N., Sung, S. H., & Hovden, R. (2020) · Microscopy and Microanalysis · doi:10.1017/s1431927620001841 · Cited by 10
“…tasks (LeCun et al., 2015), such as classification (Krizhevsky et al., 2012), encoding and decoding (Badrinarayanan et al., 2017), and regression (Mahendran et al., 2017; Lathuiliére et al., 2019); including recent applications in electron microscopy (Xu & LeBeau, 2018; Ede & Beanland, 2019;…”
BackgroundΜαρούγκας, Ι., Koutras, P., Kardaris, N., Retsinas, G., Chalvatzaki, G., & Maragos, P. A. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-66096-3_45 · Cited by 12
Kubeyev, A. (2023) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2023.106562 · Cited by 9
mmBaT: A Multi-Task Framework for Mmwave-Based Human Body Reconstruction and Translation Prediction
Yang, J., Xia, S., Song, Y., Wu, Q., & Pei, L. (2024) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp48485.2024.10448164 · Cited by 9
“The losses associated with these parameters are defined as: where L β , L γ , L J and L M are the L1 loss with the same definition in Eq.4, while L θ is the geodesic loss [19] to represent the distance between the predicted and ground-truth rotation matrices.”
BackgroundLearning Cross-View Consistent 3D Keypoints for Object 6D Pose Estimation
Zhang, S., Zhao, W., Guan, Z., Zhao, W. Z., Peng, J., & Fan, J. (2025) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2025.3538018 · Cited by 11
Chen, M., Zhang, Z., Gu, S., & Kong, Y. (2024) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp48485.2024.10446096 · Cited by 13
“The training strategy of this module is based on geodesic gradient which has been studied for registration problems due to its convexity [25, 26], and double backward mechanism which means computing parameter gradient by using a error function in first back-propagation and updating the network…”
MethodMonocular Viewpoints Estimation for Generic Objects in the Wild
Li, Z., Wang, Y., & Ji, X. (2019) · IEEE Access · doi:10.1109/access.2019.2923436 · Cited by 8
Cutting Pose Prediction from Point Clouds
Philipsen, M. P., & Moeslund, T. B. (2020) · Sensors · doi:10.3390/s20061563 · Cited by 8
“Recently, pose prediction methods have begun to break free from the restrictions of the image plane and instead represent pose as position and either Euler angles [37], quaternions [37,38], or normal vectors [29].”
BackgroundGiefer, L. A., Castellanos, J. D. A., Faghihabdolahi, M., & Freitag, M. (2020) · Procedia CIRP · doi:10.1016/j.procir.2020.05.092 · Cited by 9
Metadata
- Authors
- Mahendran, S., Ali, H., & Vidal, R.
- Affiliations
- Johns Hopkins University (US)
- ORCID
- Ali, Haider
- Venue
- 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)
- Publisher
- IEEE
- Published
- 2017-07
- Pages
- 494-495
- OpenAlex
- W2737704667
- Cited by
- 210
- References
- 24
- Citations by year
- 2026: 6 · 2025: 9 · 2024: 20 · 2023: 24 · 2022: 23 · 2021: 33 · 2020: 45 · 2019: 32 · 2018: 17 · 2017: 1
- FWCI
- 7.60
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) (publishedVersion)
- IEEE International Conference on Computer Vision Workshops (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- National Science Foundation (1527340, 1053575, 1445606)
- Maryland Advanced Research Computing Center
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Engineering › Control and Systems Engineering › Robot Manipulation and Learning
- Physical Sciences › Computer Science › Computer Graphics and Computer-Aided Design › 3D Shape Modeling and Analysis
- Keywords
- 3D pose estimation, object pose estimation, data augmentation, loss function design
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
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.