DTAM: Dense tracking and mapping in real-time
The DOI matches its Crossref record; 16 of the 17 references it cites were found in a registry or an index.
Abstract
DTAM is a system for real-time camera tracking and reconstruction which relies not on feature extraction but dense, every pixel methods. As a single hand-held RGB camera flies over a static scene, we estimate detailed textured depth maps at selected keyframes to produce a surface patchwork with millions of vertices. We use the hundreds of images available in a video stream to improve the quality of a simple photometric data term, and minimise a global spatially regularised energy functional in a novel non-convex optimisation framework. Interleaved, we track the camera's 6DOF motion precisely by frame-rate whole image alignment against the entire dense model. Our algorithms are highly parallelisable throughout and DTAM achieves real-time performance using current commodity GPU hardware. We demonstrate that a dense model permits superior tracking performance under rapid motion compared to a state of the art method using features; and also show the additional usefulness of the dense model for real-time scene interaction in a physics-enhanced augmented reality application.
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
“However, in our network, multi-scale projections not only help to overcome gradient locality during training (see [33], [47] for a broader discussion of this well-known issue), but also aid in online error correction at runtime.”
From Unsupervised Deep Visual-Inertial Odometry with Online Error Correction for RGB-D Imagery (2020) · cites this paper as [33]
In the citing paper
“SFMLearner [33] demonstrated the unsupervised learning of unscaled egomotion and depth from RGB imagery.”
From Unsupervised Deep Visual-Inertial Odometry with Online Error Correction for RGB-D Imagery (2020) · cites this paper as [33]
In the citing paper
“For KITTI, VIOLearner RGB outperforms approaches SFMLearner [33] and the network of Zhan et al. [30].”
From Unsupervised Deep Visual-Inertial Odometry with Online Error Correction for RGB-D Imagery (2020) · cites this paper as [33]
In the citing paper
“For AirSim, VIOLearner (no IMU) outperforms approaches SFMLearner [33] and ORB-SLAM2-S in translation error.”
From Unsupervised Deep Visual-Inertial Odometry with Online Error Correction for RGB-D Imagery (2020) · cites this paper as [33]
In the citing paper
“Compared to other vision-only approaches, for KITTI, VIOLearner (no IMU) outperforms approaches SFMLearner [33] and the network of Zhan et al. [30] and yet is outperformed by ORB-SLAM-2 which uses full bundle adjustment.”
From Unsupervised Deep Visual-Inertial Odometry with Online Error Correction for RGB-D Imagery (2020) · cites this paper as [33]
Citations
Cites
Nonlinear total variation based noise removal algorithms
Rudin, L. I., Osher, S., & Fatemi, E. (1992) · Physica D: Nonlinear Phenomena · doi:10.1016/0167-2789(92)90242-f · Cited by 15,704
Parallel 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
A First-Order Primal-Dual Algorithm for Convex Problems with Applications to Imaging
Chambolle, A., & Pock, T. (2011) · Journal of Mathematical Imaging and Vision · doi:10.1007/s10851-010-0251-1 · Cited by 4,336
Lucas-Kanade 20 Years On: A Unifying Framework
Baker, S., & Matthews, I. (2004) · International Journal of Computer Vision · doi:10.1023/b:visi.0000011205.11775.fd · Cited by 3,155
Fast cost-volume filtering for visual correspondence and beyond
Rhemann, C., Hosni, A., Bleyer, M., Rother, C., & Gelautz, M. (2011) · CVPR 2011 · doi:10.1109/cvpr.2011.5995372 · Cited by 485
Live dense reconstruction with a single moving camera
Newcombe, R. A., & Davison, A. J. (2010) · 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2010.5539794 · Cited by 366
Self-citationImproving the Agility of Keyframe-Based SLAM
Klein, G., & Murray, D. (2008) · Lecture Notes in Computer Science · doi:10.1007/978-3-540-88688-4_59 · Cited by 335
Robust multi-sensor image alignment
Irani, M., & Anandan, P. (2002) · Sixth International Conference on Computer Vision (IEEE Cat. No.98CH36271) · doi:10.1109/iccv.1998.710832 · Cited by 220
Real-Time Dense Geometry from a Handheld Camera
Stühmer, J., Gumhold, S., & Cremers, D. (2010) · Lecture Notes in Computer Science · doi:10.1007/978-3-642-15986-2_2 · Cited by 205
Some First-Order Algorithms for Total Variation Based Image Restoration
Aujol, J. (2009) · Journal of Mathematical Imaging and Vision · doi:10.1007/s10851-009-0149-y · Cited by 151
Cited by
Items 601–700 of 1,004
Unsupervised Depth Learning in Challenging Indoor Video: Weak Rectification to Rescue.Preprint
Bian, J.-W., Zhan, H., Wang, N., Chin, T.-J., Shen, C., & Reid, I. D. (2020) · arXiv (Cornell University) · Cited by 26
MAMo: Leveraging Memory and Attention for Monocular Video Depth Estimation
Yasarla, R., Cai, H., Jeong, J., Shi, Y., Garrepalli, R., & Porikli, F. (2023) · 2023 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00804 · Cited by 15
“Traditional approaches [47, 14, 39] utilize stereo vision and/or structure-from-motion to estimate depth, which, however, have limited accuracy.”
BackgroundORB-NeuroSLAM: A Brain-Inspired 3-D SLAM System Based on ORB Features
Shen, D., Liu, G., Li, T., Yu, F., Gu, F., Xiao, K., & Zhu, X. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3335417 · Cited by 29
“Newcombe et al. [18] introduced the first direct vSLAM method, called DTAM.”
MethodHierarchical Higher-Order Regression Forest Fields: An Application to 3D Indoor Scene Labelling
Pham, T. T., Reid, I. D., Latif, Y., & Gould, S. J. (2015) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.259 · Cited by 25
“In recent years significant progress has been made in Structure from Motion (SfM) and Visual Simultaneous Localisation and Mapping (VSLAM) and it is now possible to perform real-time, dense 3D reconstruction in an indoor environment using either a single camera [18] or RGB-D sensor [17].”
BackgroundClement, L., & Kelly, J. W. (2018) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2018.2799741 · Cited by 29
“Following the advice of [1], we optimize only for rotation at the lowest resolution.”
BackgroundMethodDeep Visual Odometry With Adaptive Memory
Xue, F., Wang, X., Wang, J., & Zha, H. (2020) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2020.3014100 · Cited by 29
“As an essential task in various applications such as autonomous driving, virtual/augmented reality, and robot navigation, VO has been studied for decades and many outstanding algorithms have been developed from the aspect of geometry [1], [2], [3], [4], [5], [6], [7], [8].”
BackgroundMethodZhao, Y., Chen, L., Zhang, X., Xu, S., Bu, S., Jiang, H., Han, P., Li, K., & Wan, G. (2021) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2021.3090203 · Cited by 32
“In recent years, several keypoint-based [17]–[19], direct [20]–[22], and semidirect monocular visual SLAM systems [23], [24] have been proposed and developed rapidly.”
BackgroundMethodUnderwater Exploration and Mapping
Joshi, B., Xanthidis, M., Roznere, M., Burgdorfer, N., Mordohai, P., Li, A. Q., & Rekleitis, I. M. (2022) · IEEE/OES Autonomous Underwater Vehicles · doi:10.1109/auv53081.2022.9965805 · Cited by 35
“Implicit representations that aim to compute a function whose zero-level set is the surface have been employed by conventional systems [70], [71] yielding high-quality models in real time, but are limited to small volumes.”
BackgroundDesign and Evaluation of a Generic Visual SLAM Framework for Multi Camera Systems
Kaveti, P., Vaidyanathan, S. N., Chelvan, A. T., & Singh, H. (2023) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2023.3316609 · Cited by 36
“Most VSLAM pipelines [1], [2], [3], [4] use monocular cameras prone to a single point of failure due to degenerate motion, illumination changes, featureless surfaces, and dynamic objects [5].”
BackgroundConcurrent filtering and smoothing
Kaess, M., Williams, S. R., Indelman, V., Roberts, R. B., Leonard, J. J., & Dellaert, F. (2012) · DSpace@MIT (Massachusetts Institute of Technology) · Cited by 30
Multi-view Stereo with Single-View Semantic Mesh Refinement
Romanoni, A., Ciccone, M., Visin, F., & Matteucci, M. (2017) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw.2017.89 · Cited by 30
Building maps for autonomous navigation using sparse visual SLAM features
Ling, Y., & Shen, S. (2017) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2017.8202316 · Cited by 26
“In contrast to fine-grained 3D reconstruction methods which aim to recover detailed surface structures [1], autonomous navigation requires 3D maps that are scalable to large-scale environments and can be reconstructed incrementally in real-time.”
BackgroundIllumination Robust Monocular Direct Visual Odometry for Outdoor Environment Mapping
Wu, X., & Pradalier, C. (2019) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2019.8793607 · Cited by 25
“However, recent works have shown that the direct method [4] [5] [6] [7] could provide more accurate and robust motion estimation due to their high flexibility of image information usage in comparison of the indirect approaches using certain types of features.”
BackgroundQuadtree-Accelerated Real-Time Monocular Dense Mapping
Wang, K., Ding, W., & Shen, S. (2018) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2018.8594101 · Cited by 26
“DTAM [8], VI-MEAN [9] and REMODE [10] are methods that solves the dense mapping problem using global or semi-global information.”
BackgroundMethodSparse depth sensing for resource-constrained robots
Ma, F., Carlone, L., Ayaz, U., & Karaman, S. (2019) · The International Journal of Robotics Research · doi:10.1177/0278364919850296 · Cited by 30
“A variety of approaches has been proposed to perform geometric reconstruction using these sensors, for instance see [7–9] and the references therein.”
Lightweight Unsupervised Deep Loop ClosurePreprint
Merrill, N., & Huang, G. P. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1805.07703 · Cited by 25
“Due to its lightweight yet robust design, our model is suitable for use in real-time SLAM systems – in particular, direct algorithms [48, 25, 49, 50] where no intermediate image representation is needed.”
MethodSCFusion: Real-time Incremental Scene Reconstruction with Semantic Completion
Wu, S. ;. (n.d.) · mediaTUM (Technical University of Munich) · Cited by 25
Single-view 3D Scene Reconstruction with High-fidelity Shape and Texture
Chen, Y., Ni, J., Jiang, N., Zhang, Y., Zhu, Y., & Huang, S. (2024) · International Conference on 3D Vision (3DV) · doi:10.1109/3dv62453.2024.00142 · Cited by 48
“However, due to shape-appearance ambiguity, simply incorporating rendering supervision may lead to generating realistic textured images but inconsistent underlying geometry [12, 51], especially under partial observations.”
BackgroundA dense semantic mapping system based on CRF-RNN network
Cheng, J., Sun, Y., & Meng, M. Q. ‐. (2017) · International Conference on Advanced Robotics (ICAR) · doi:10.1109/icar.2017.8023671 · Cited by 24
SLAM-Based Follow-the-Leader Deployment of Concentric Tube Robots
Girerd, C., Kudryavtsev, A. V., Rougeot, P., Renaud, P., Rabenorosoa, K., & Tamadazte, B. (2020) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2019.2963821 · Cited by 30
“[21] R. A. Newcombe, S. J. Lovegrove, and A. J. Davison, “DTAM: Dense tracking and mapping in real-time,” in Proc.”
MethodRecovering dense 3D point clouds from single endoscopic image
Xi, L., Zhao, Y., Chen, L., Gao, Q.-H., Tang, W., Wan, T.-R., & Xue, T. (2021) · Comput. Methods Programs Biomed. · doi:10.1016/j.cmpb.2021.106077 · Cited by 24
RGB-Fusion: Monocular 3D reconstruction with learned depth prediction
Duan, Z., Chen, Y., Yu, H., Hu, B., & Chen, C. (2021) · Displays · doi:10.1016/j.displa.2021.102100 · Cited by 28
Stereo Plane SLAM Based on Intersecting Lines
Zhang, X., Wang, W., Qi, X., & Liao, Z. (2021) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros51168.2021.9635961 · Cited by 24
“Point features can be used to build a sparse map [13], semi-dense map [14] and even dense map [15].”
BackgroundSemi-direct visual odometry for a fisheye-stereo camera
Heng, L., & Choi, B. (2016) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2016.7759600 · Cited by 28
“In contrast, direct visual odometry methods [16] attempt per-pixel depth estimation, and thus, require the use of a commodity GPU for real-time”
BackgroundPLI-VINS: Visual-Inertial SLAM Based on Point-Line Feature Fusion in Indoor Environment
Zhao, Z., Song, T., Xing, B., Yu, L., & Wang, Z. (2022) · Sensors · doi:10.3390/s22145457 · Cited by 29
“Direct methods, such as DTAM [3], LSDSLAM [4] and DSO [5], estimate camera motion according to the pixel brightness information of the image, and optimize the solution by minimizing the brightness error.”
MethodDNS-SLAM: Dense Neural Semantic-Informed SLAM
Li, K., Niemeyer, M., Navab, N., & Tombari, F. (2024) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros58592.2024.10803056 · Cited by 35
“In summary, our main contributions include: • Leveraging 2D semantic priors, we investigate a semantic-informed multi-class scene representation, yielding an efficient, comprehensive, and semantically decomposed geometry representation.”
HDRFusion: HDR SLAM Using a Low-Cost Auto-Exposure RGB-D Sensor
Li, S., Handa, A., Zhang, Y., & Calway, A. D. (2016) · International Conference on 3D Vision (3DV) · doi:10.1109/3dv.2016.40 · Cited by 24
“The former assumes that any two overlapping frames from an input RGB sequence fulfil the condition [20], while the latter requires only that consecutive frames do[12, 25].”
BackgroundMethodInfluentialRobust Monocular Visual Teach and Repeat Aided by Local Ground Planarity and Color‐constant Imagery
Clement, L., Kelly, J. W., & Barfoot, T. D. (2016) · Journal of Field Robotics · doi:10.1002/rob.21655 · Cited by 27
“Recently, dense photometric approaches to monocular SLAM have been developed that operate directly on per-pixel intensity measurements to produce maps with a higher spatial density than is typically achievable using sparse keypoint-based methods (Newcombe et al., 2011; Engel et al., 2014; Pizzoli et al., 2014).”
MethodBrock, O., Park, J., & Toussaint, M. (2016) · Springer handbooks · doi:10.1007/978-3-319-32552-1_40 · Cited by 27
Towards real-time underwater 3D reconstruction with plenoptic cameras
Skinner, K. A., & Johnson‐Roberson, M. (2016) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2016.7759317 · Cited by 26
“Recently developed RGB-D simultaneous localization and mapping (SLAM) systems have demonstrated the ability to perform this task with a high degree of success for land-based applications [6] [7].”
BackgroundAutostereoscopic augmented reality visualization for depth perception in endoscopic surgery
Wang, R., Geng, Z., Zhang, Z., Pei, R., & Meng, X. (2017) · Displays · doi:10.1016/j.displa.2017.03.003 · Cited by 24
Distributed monocular visual SLAM as a basis for a collaborative augmented reality framework
Egodagamage, R., & Tüceryan, M. (2018) · Computers & Graphics · doi:10.1016/j.cag.2018.01.002 · Cited by 24
“[11] Newcombe RA, Lovegrove S, Davison A.”
Real-Time Underwater StereoFusion
Rossi, M., Trslić, P., Sivčev, S., Riordan, J. F., Toal, D., & Dooly, G. (2018) · Sensors · doi:10.3390/s18113936 · Cited by 34
“The first system to use both dense tracking and mapping and capable of real-time processing was presented by [27].”
MethodUsing Unsupervised Deep Learning Technique for Monocular Visual Odometry
Liu, Q., Li, R., Hu, H., & Gu, D. (2019) · IEEE Access · doi:10.1109/access.2019.2896988 · Cited by 23
“[26], [27], namely dense tracking and mapping (DTAM), which can be viewed as a direct method.”
BackgroundMask R-CNN Based Semantic RGB-D SLAM for Dynamic Scenes
Zhang, Z., Zhang, J., & Tang, Q. (2019) · IEEE/ASME International Conference on Advanced Intelligent Mechatronics · doi:10.1109/aim.2019.8868400 · Cited by 26
“The first group, direct methods, can build a completely dense reconstruction and accurately estimate camera motion by the direct minimization of the photometric error between two adjacent images [7].”
BackgroundAlamdari, A. G., Zade, F. A., & Ebrahimkhanlou, A. (2025) · Sensors · doi:10.3390/s25030712 · Cited by 37
Unsupervised Dense Object Discovery, Detection, Tracking and Reconstruction
Ma, L., & Sibley, G. (2014) · Lecture notes in computer science · doi:10.1007/978-3-319-10605-2_6 · Cited by 22
“In future work, we aim to improve the resolution of the reconstructed models by [12,23] and enhanced the system with a single RGB camera techniques from [15,13].”
MethodResultOnline Reconstruction of Indoor Scenes from RGB-D Streams
Wang, H., Wang, J., & Wang, L. (2016) · 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.2016.356 · Cited by 23
“Attempts at real-time dense surface modelling from passive cameras to date have had limited results, most- ∗indicates equal contributions and joint first authors ly on outdoor scenes [23, 30], small objects and compact workspaces [16, 18, 24].”
BackgroundMeaningful maps with object-oriented semantic mapping
Sünderhauf, N., Pham, T. T., Latif, Y., Milford, M. J., & Reid, I. D. (2017) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2017.8206392 · Cited by 126
“Pham et al. [8] first reconstruct a dense 3D model from RGB-D images using KinectFusion [9], then assign every 3D point a semantic label using a hierarchical Conditional Random Fields (CRF) model.”
MethodReal-time local 3D reconstruction for aerial inspection using superpixel expansion
Teixeira, L., & Chli, M. (2017) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2017.7989530 · Cited by 24
“Shortly after, the emergence of KinectFusion, came DTAM [9] to pioneer online dense reconstruction using a monocular camera, achieving comparable reconstruction quality to KinectFusion.”
MethodDense monocular reconstruction using surface normals
Weerasekera, C. S., Latif, Y., Garg, R., & Reid, I. D. (2017) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2017.7989293 · Cited by 24
“We extensively evaluate the proposed method quantitatively against the traditional hand-crafted smoothness regularizer such as [4], and pure learning-based system like [12] on a diverse range of indoor sequences in a large dataset like raw NYU-D V2 [16].”
BackgroundMethodInfluential3D plant root system reconstruction based on fusion of deep structure-from-motion and IMU
Lu, Y., Wang, Y., Chen, Z., Khan, A., Salvaggio, C., & Lu, G. (2021) · Multimedia Tools and Applications · doi:10.1007/s11042-020-10069-3 · Cited by 27
“Classical structure from motion (SfM) methods refer to a process of 3D reconstruction and camera pose estimation from multiple ordered or unordered images from different perspectives [1, 2, 24, 42, 51].”
BackgroundKrig, S. (2016) · Springer eBooks · doi:10.1007/978-3-319-33762-3_2 · Cited by 32
Multi-Modal Neural Radiance Field for Monocular Dense SLAM with a Light-Weight ToF Sensor
Liu, X., Li, Y., Teng, Y., Bao, H., Zhang, G., Zhang, Y., & Cui, Z. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00007 · Cited by 23
“In all our experiments, we set the grid sizes to [3, 6, 24, 96] cm and the number of sampling pixels M to 5000.”
A Fast Robot Identification and Mapping Algorithm Based on Kinect Sensor
Zhang, L., Shen, P., Zhu, G., Wei, W., & Song, H. (2015) · Sensors · doi:10.3390/s150819937 · Cited by 24
“The improved RGB-D SLAM [17] has great efficiency and high precision.”
BackgroundOn the Accuracy of Dense Fisheye Stereo
Schneider, J., Stachniss, C., & Förstner, W. (2016) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2016.2516509 · Cited by 23
“The third subsection describes how we compute the dense 3D point cloud with its uncertainty through variance propagation using the disparity information.”
MethodMonoRGBD-SLAM: Simultaneous localization and mapping using both monocular and RGBD cameras
Yousif, K., Taguchi, Y., & Ramalingam, S. (2017) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2017.7989521 · Cited by 24
“Most recently, due to the increase in computational capability, direct methods such as DTAM [6] and LSD-SLAM [1] have been proposed.”
MethodFLaME: Fast Lightweight Mesh Estimation Using Variational Smoothing on Delaunay Graphs
Greene, W. N., & Roy, N. (2017) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.502 · Cited by 33
Stereo Visual-Inertial SLAM With Points and Lines
Liu, Y., Yang, D., Li, J., Gu, Y., Pi, J., & Zhang, X. (2018) · IEEE Access · doi:10.1109/access.2018.2880689 · Cited by 24
“DTAM [24] computes a dense depth map from a single moving camera in real time.”
BackgroundEvent-based, 6-DOF Camera Tracking for High-Speed Applications.Preprint
Gallego, G., Lund, J. E. A., Mueggler, E., Rebecq, H., Delbrück, T., & Scaramuzza, D. (2016) · arXiv (Cornell University) · Cited by 23
Video Based Reconstruction of 3D People ModelsPreprint
Alldieck, T., Magnor, M. A., Xu, W., Theobalt, C., & Pons‐Moll, G. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1803.04758 · Cited by 47
“Alternatively, multiview passive reconstruction from a dense set of static body pose images can be used [22, 46].”
BackgroundMoD-SLAM: Monocular Dense Mapping for Unbounded 3D Scene Reconstruction
Zhou, H., Guo, Z., Ren, Y., Liu, S., Zhang, L., Zhang, K., & Li, M. (2024) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2024.3498777 · Cited by 27
HE-SLAM: A Stereo SLAM System Based on Histogram Equalization and ORB Features
Fang, Y., Shan, G., Wang, T., Li, X., Liu, W., & Snoussi, H. (2018) · Chinese Automation Congress · doi:10.1109/cac.2018.8623424 · Cited by 26
“The direct methods (typical algorithms such as DTAM [4], LSD-SLAM [5], and DSO [6], etc.) estimate the camera motion according to the pixel luminance information of the image and optimizes it by minimizing the luminosity error.”
MethodArea-based correlation and non-local attention network for stereo matching
Li, X., Fan, Y., Lv, G., & Ma, H. (2021) · The Visual Computer · doi:10.1007/s00371-021-02228-w · Cited by 25
“Although in principle, a simple argmin on the cost volume should lead to a correct local minimum solution, the literature [16,30,39] has shown that it is common for a solution to have multiple local minimums.”
BackgroundTightCap: 3D Human Shape Capture with Clothing Tightness Field
Chen, X., Pang, A., Yang, W., Wang, P., Xu, L., & Yu, J. (2021) · ACM Transactions on Graphics · doi:10.1145/3478518 · Cited by 47
“The former approaches [Collet et al. 2015; Furukawa et al. 2013; Newcombe et al. 2011b; Strecha et al. 2008] employ correspondence matching and triangulation.”
Zhang, H., Lin, Y., Teng, F., & Hong, W. (2022) · Remote Sensing · doi:10.3390/rs14225715 · Cited by 26
Luó, X., & Mori, K. (2013) · IEEE Transactions on Biomedical Engineering · doi:10.1109/tbme.2013.2277609 · Cited by 22
“Finally, our proposed APF method for endoscope tracking and navigation is significantly related to visual simultaneous localization and mapping (SLAM) [21], [22].”
MethodFlexible three-dimensional modeling of plants using low- resolution cameras and visual odometry
Santos, T. T., & Rodrigues, G. C. (2015) · Machine Vision and Applications · doi:10.1007/s00138-015-0729-3 · Cited by 25
“Recently, novel systems for dense three-dimensional reconstruction in real time were proposed [42,46].”
BackgroundMethodFeature Learning and Deep Learning Architecture Survey
Krig, S. (2016) · Springer eBooks · doi:10.1007/978-3-319-33762-3_10 · Cited by 25
Direct methods for 3D reconstruction and visual SLAM
Cremers, D. (2017) · Machine Vision Applications · doi:10.23919/mva.2017.7986766 · Cited by 22
“The work [18] tackles the joint problem of camera motion estimation and structure reconstruction by means of a direct method, i.”
BackgroundMethodAccurate Direct Visual-Laser Odometry with Explicit Occlusion Handling and Plane Detection
Huang, K., Xiao, J., & Stachniss, C. (2019) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2019.8793629 · Cited by 25
“Recent advantages in this approach showed remarkable performance [3], [5], [9], [13], [20].”
BackgroundTerrainFusion: Real-time Digital Surface Model Reconstruction based on Monocular SLAM
Wang, W., Zhao, Y., Han, P., Zhao, P., & Bu, S. (2019) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros40897.2019.8967663 · Cited by 28
“SLAM technology develops rapidly and a variety of SLAM systems have been proposed, including monocular SLAM system (key-point based [34], [35], [21], direct [36], [22], [37], and semi-direct methods [23], [38]), multi-sensor SLAM system (RGBD [39], [40], [41], Stereo [42], [38], [43] and inertial”
BackgroundLee, H., Chun, J., & Jeon, K. (2020) · IEEE Transactions on Aerospace and Electronic Systems · doi:10.1109/taes.2020.3016873 · Cited by 22
Scalable structure from motion for densely sampled videos
Resch, B., Lensch, H. P. A., Wang, O., Pollefeys, M., & Sorkine‐Hornung, A. (2015) · 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.2015.7299019 · Cited by 23
“Real-time methods based on feature points [9] or dense, per-pixel tracking [21] are generally designed to provide as good as possible results with a small input lag, rather than a final, fully consistent and high quality reconstruction that globally optimizes the poses of all input frames.”
MethodCGS-SLAM: Compact 3D Gaussian Splatting for Dense Visual SLAM
Deng, T., Chen, Y., Yang, J., Yuan, S., Liu, J., Wang, D., & Chen, W. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11246845 · Cited by 35
“Attention has turned to dense scene reconstruction, exemplified by DTAM [9], Kintinuous [10], and ElasticFusion [11].”
Agudo, A., Montiel, J. M. M., Agapito, L., & Calvo, B. (2016) · Journal of Mathematical Imaging and Vision · doi:10.1007/s10851-016-0668-2 · Cited by 21
Learning Meshes for Dense Visual SLAM
Bloesch, M., Laidlow, T., Clark, R., Leutenegger, S., & Davison, A. J. (2019) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00595 · Cited by 28
“To account for this, a regulariser is typically added, often based on smoothness [23] or planar [5] assumptions.”
BackgroundMethodInfluentialSelf-citationConstant Velocity Constraints for Self-Supervised Monocular Depth Estimation
Zhou, H., Greenwood, D., Taylor, S. L., & Gong, H. (2020) · Conference on Visual Media Production · doi:10.1145/3429341.3429355 · Cited by 22
Joint Optimization of Depth and Ego-Motion for Intelligent Autonomous Vehicles
Gao, Y., Tian, F., Li, J., Fang, Z., Al–Rubaye, S., Song, W., & Yan, Y. (2022) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2022.3159275 · Cited by 24
“DTAM [6] and REMODE [12] generally create dense scenes by optimizing depth values.”
BackgroundMethodVidSfM: Robust and Accurate Structure-From-Motion for Monocular Videos
Cui, H., Tu, D., Tang, F., Xu, P., Liu, H., & Shen, S. (2022) · IEEE Transactions on Image Processing · doi:10.1109/tip.2022.3156375 · Cited by 22
Learning a Depth Covariance Function
Dexheimer, E., & Davison, A. J. (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/cvpr52729.2023.01261 · Cited by 24
“View-based Priors for Geometric Inference Traditionally, sparse reconstruction methods have ignored per-image correlations [39], [28], while dense methods [29], [13] have used naive priors based on neighboring pixels.”
MethodSelf-citationZhang, Z., Chan, R. K. Y., & Wong, K. K. Y. (2023) · Neurocomputing · doi:10.1016/j.neucom.2023.127122 · Cited by 24
Development of vision–based SLAM: from traditional methods to multimodal fusion
Zheng, Z., Su, K., Lin, S., Fu, Z., & Yang, C. (2024) · Robotic Intelligence and Automation · doi:10.1108/ria-10-2023-0142 · Cited by 29
Oskiper, T., Sizintsev, M., Branzoi, V., Samarasekera, S., & Kumar, R. (2015) · IEEE Transactions on Visualization and Computer Graphics · doi:10.1109/tvcg.2015.2408612 · Cited by 25
“…systems concentrated on marker-based augmented reality for speed, simplicity and low hardware requirements (e.g., [6], [19], [20]), a number of good marker-less vision-only pose tracking systems that use natural image features/lines/edges appeared over the past decade (e.g., [4], [7], [12], [17]).”
Cost-aware depth map estimation for Lytro camera
Kim, M., Oh, T.-H., & Kweon, I. S. (2014) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2014.7025006 · Cited by 21
Variational PatchMatch MultiView Reconstruction and Refinement
Heise, P., Jensen, B., Klose, S., & Knoll, A. (2015) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.107 · Cited by 26
“To refine the initial poses we use direct image alignment as commonly used for Visual Odometry estimation using RGB-D sensors [34, 30, 21].”
MethodInfluentialLarge-scale, real-time 3D scene reconstruction on a mobile device
Dryanovski, I., Klingensmith, M., Srinivasa, S. S., & Xiao, J. (2017) · Autonomous Robots · doi:10.1007/s10514-017-9624-2 · Cited by 23
“Newcombe et al. (2011) uses a TSDF to simultaneously extract the pose of a moving depth camera and scene geometry in real-time.”
BackgroundMethodInfluentialRepresentation of truss-style structures for autonomous climbing of biped pole-climbing robots
Chen, W., Gu, S., Zhu, L., Zhang, H., Zhu, H., & Guan, Y. (2018) · Robotics and Autonomous Systems · doi:10.1016/j.robot.2018.01.002 · Cited by 21
Edge-Based Robust RGB-D Visual Odometry Using 2-D Edge Divergence Minimization
Kim, C., Kim, P., Lee, S.-I., & Kim, H. J. (2018) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2018.8593594 · Cited by 22
“Camera motion estimation from consecutive images, known as visual odometry (VO) [1], is receiving increasing attention in areas of autonomous robots [13] and virtual and augmented reality (VR/AR) applications requiring selflocalization abilities [3], [4].”
BackgroundOverview obstacle maps for obstacle‐aware navigation of autonomous drones
Pestana, J., Maurer, M., Muschick, D., Hofer, M., & Fraundorfer, F. (2019) · Journal of Field Robotics · doi:10.1002/rob.21863 · Cited by 21
“…(Newcombe, Izadi, et al., 2011), ElasticFusion (Whelan, Leute-negger, Salas ‐ Moreno, Glocker, & Davison, 2015, 2016), MonoFusion (Pradeep et al., 2013), and dense tracking and mapping (DTAM; Newcombe, Lovegrove, & Davison, 2011), or the similar approach by Stühmer, Gumhold, and Cremers (2010).”
ResultThe role of prior in image based 3D modeling: a survey
Zhu, H., Nie, Y., Yue, T., & Cao, X. (2016) · Frontiers of Computer Science · doi:10.1007/s11704-016-5520-8 · Cited by 21
“The state-of-theart SFM systems can achieve real-time and dense tracking SLAM using inexpensive cameras [12,13], even for the dynamic scenes [14].”
BackgroundRelative Camera Refinement for Accurate Dense Reconstruction
Yao, Y., Li, S., Zhu, S., Deng, H., Fang, T., & Quan, L. (2017) · International Conference on 3D Vision (3DV) · doi:10.1109/3dv.2017.00030 · Cited by 22
“Apart from the offline SfM methods, in robotic areas researchers apply the SLAM methods to reconstruct the camera odometry in real-time [10, 21, 30, 32, 12].”
MethodDitto in the House: Building Articulation Models of Indoor Scenes through Interactive Perception
Hsu, C.-C., Jiang, Z., & Zhu, Y. (2023) · 2023 IEEE International Conference on Robotics and Automation (ICRA) · doi:10.1109/icra48891.2023.10161431 · Cited by 21
FC-SLAM: Federated Learning Enhanced Distributed Visual-LiDAR SLAM In Cloud Robotic System
Li, Z., Wang, L., Jiang, L., & Xu, C.-Z. (2019) · 2019 IEEE International Conference on Robotics and Biomimetics (ROBIO) · doi:10.1109/robio49542.2019.8961798 · Cited by 28
“Studies in recent years include DTAM [12], DSO [13] and so on.”
BackgroundDNA-SLAM: Dense Noise Aware SLAM for ToF RGB-D Cameras
Wasenmüller, O., Ansari, M. D., & Stricker, D. (2016) · ACCV Workshops · doi:10.1007/978-3-319-54407-6_42 · Cited by 21
VOLDOR+SLAM: For the times when feature-based or direct methods are not good enough
Min, Z., & Dunn, E. (2021) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48506.2021.9561230 · Cited by 30
“Dense direct methods [4] are typically sensitive to illumination changes, and need a sufficient number of images to triangulate structure at poorly”
BackgroundHybrid sparse monocular visual odometry with online photometric calibration
Luo, D., Zhuang, Y., & Wang, S. (2022) · The International Journal of Robotics Research · doi:10.1177/02783649221107703 · Cited by 21
“Direct Methods: Direct methods (Engel et al., 2013, 2018; Newcombe et al., 2011) which make an assumption on brightness constancy do not rely on the feature extraction.”
MethodStereo Visual Odometry With Automatic Brightness Adjustment and Feature Tracking Prediction
Yin, H., Liu, P. X., & Zheng, M. (2022) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2022.3223070 · Cited by 19
EGA-Depth: Efficient Guided Attention for Self-Supervised Multi-Camera Depth Estimation
Shi, Y., Cai, H., Ansari, A., & Porikli, F. (2023) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw59228.2023.00017 · Cited by 26
“Traditional approaches [15, 35, 39] utilize stereoscopic vision and/or structure-from-motion to estimate depth.”
BackgroundVisual SLAM Combining Lines and Structural Regularities: Towards Robust Localization
Zhao, W., Sun, H., Zhang, X., & Xiong, Y. (2023) · IEEE Transactions on Intelligent Vehicles · doi:10.1109/tiv.2023.3311511 · Cited by 27
“Notably, the reconstruction from the direct method based on feature points, partial, or full pixels can be sparse, semi-dense, or dense [21], [22].”
MethodMGS-SLAM: Monocular Sparse Tracking and Gaussian Mapping With Depth Smooth Regularization
Zhu, P., Zhuang, Y., Chen, B., Li, L., Wu, C., & Liu, Z. (2024) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2024.3460431 · Cited by 27
“DTAM [5] pioneered one of the earliest real-time dense SLAM systems by performing parallel depth computations on GPU.”
BackgroundIncorporating scene priors to dense monocular mapping
Concha, A., Hussain, W., Montano, L., & Civera, J. (2015) · Autonomous Robots · doi:10.1007/s10514-015-9465-9 · Cited by 20
“…12 is then minimized as follows ˆ ρ = arg max q , || q || 2 ≤ 1 arg min ρ , a E ( ρ , a , q ) (13) Where q is the dual variable, A ρ stands for the gradient of ρ , (cid:15) is the threshold of the Huber norm which determines when L 1 or L 2 norm are used [28] and δ q is an indicator function [18].”
BackgroundMethodResultInfluentialChen, L., Wang, S., Hu, H., Gu, D., & Liao, L. (2015) · IEEE Sensors Journal · doi:10.1109/jsen.2015.2432082 · Cited by 23
“However, most of the research has been focused for terrestrial [5], [6] or flying robot [7], [8], and very few are focused on underwater environments.”
BackgroundSimultaneous localization, mapping, and manipulation for unsupervised object discovery
Ma, L., Ghafarianzadeh, M., Coleman, D., Correll, N., & Sibley, G. (2015) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2015.7139365 · Cited by 20
“The system could use a single RGB camera by using techniques from [22], [23].”
Method2D or Not 2D: Bridging the Gap Between Tracking and Structure from Motion
Lebeda, K., Hadfield, S., & Bowden, R. (2015) · Lecture notes in computer science · doi:10.1007/978-3-319-16817-3_42 · Cited by 20
“Recently, there has been a shift from EKF to BA [33, 34] and from sparse to dense features [35], ultimately leading to whole-image-alignment works [36, 37].”
BackgroundSDICP: Semi-Dense Tracking based on Iterative Closest Points
Kneip, L., Zhou, Y. B., & Li, H. (2015) · British Machine Vision Conference (BMVC) · doi:10.5244/c.29.100 · Cited by 20
“[12] and [4] are conceptually similar to [17], however gain computational efficiency by reducing the computation from dense to sparse or semi-dense regions.”
BackgroundMethodPinpoint SLAM: A hybrid of 2D and 3D simultaneous localization and mapping for RGB-D sensors
Ataer-Cansızoğlu, E., Taguchi, Y., & Ramalingam, S. (2016) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2016.7487262 · Cited by 20
“Recent dense SLAM systems, both monocular [10], [11] and 3D/RGBD [12], [13], [14], [15], [16], do not rely on feature extraction but rather exploit all the 2D or 3D points in frames and minimize photometric errors or iterative closest point (ICP) costs for direct registration.”
BackgroundHigh-fidelity simulation for evaluating robotic vision performance
Skinner, J., Garg, S., Sünderhauf, N., Corke, P., Upcroft, B., & Milford, M. J. (2016) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2016.7759425 · Cited by 26
“dense [5], [20] or dense [16] reconstruction of environment for visual SLAM and other similar applications.”
BackgroundSDF-2-SDF Registration for Real-Time 3D Reconstruction from RGB-D Data
Slavcheva, M., Kehl, W., Navab, N., & Ilić, S. (2017) · International Journal of Computer Vision · doi:10.1007/s11263-017-1057-z · Cited by 24
“Given a trajectory estimated in this manner, a reconstruction can be generated in various ways, among which volumetric fusion [3,51], carefully selected keyframe fusion [47], or point-based fusion [36].”
BackgroundKO-Fusion: Dense Visual SLAM with Tightly-Coupled Kinematic and Odometric Tracking
Houseago, C., Bloesch, M., & Leutenegger, S. (2019) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2019.8793471 · Cited by 24
“The integration of Simultaneous Localization and Mapping (SLAM) algorithms into real-world agents is a significant challenge in the future of mobile robotics.”
BackgroundInfluential
Metadata
- Authors
- Newcombe, R. A., Lovegrove, S. J., & Davison, A. J.
- Affiliations
- Imperial College London (GB)
- ORCID
- Davison, Andrew J.
- Venue
- 2011 International Conference on Computer Vision
- Publisher
- IEEE
- Published
- 2011-11
- Pages
- 2320-2327
- OpenAlex
- W2108134361
- Cited by
- 2,078
- References
- 17
- Citations by year
- 2026: 49 · 2025: 144 · 2024: 163 · 2023: 140 · 2022: 141 · 2021: 166 · 2020: 179 · 2019: 199 · 2018: 183 · 2017: 176 · 2016: 136 · 2015: 151 · 2014: 102 · 2013: 92 · 2012: 51
- FWCI
- 154.56
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- IEEE/CVF International Conference on Computer Vision (ICCV) (publishedVersion)
- 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
- dense 3D reconstruction, real-time mapping, depth images, non-convex optimization, energy minimization, image registration, GPU parallelization, augmented reality, handheld camera
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.