Filters








35 Hits in 0.86 sec

DeepTAM: Deep Tracking and Mapping [chapter]

Huizhong Zhou, Benjamin Ummenhofer, Thomas Brox
2018 Lecture Notes in Computer Science  
We present a system for keyframe-based dense camera tracking and depth map estimation that is entirely learned.  ...  We demonstrate that the performance of our 6 DOF tracking competes with RGB-D tracking algorithms.We compare favorably against strong classic and deep learning powered dense depth algorithms.  ...  We also thank the bwHPC initiative for computing resources, Facebook for their P100 server donation and gift funding.  ... 
doi:10.1007/978-3-030-01270-0_50 fatcat:a4mjbsavszd7tlf3nszogkvcty

DeepTAM: Deep Tracking and Mapping [article]

Huizhong Zhou, Benjamin Ummenhofer, Thomas Brox
2018 arXiv   pre-print
We present a system for keyframe-based dense camera tracking and depth map estimation that is entirely learned.  ...  We demonstrate that the performance of our 6 DOF tracking competes with RGB-D tracking algorithms. We compare favorably against strong classic and deep learning powered dense depth algorithms.  ...  As a consequence, DeepTAM generalizes well to new datasets and is the first learned approach with full 6 DOF keyframe pose tracking and dense mapping.  ... 
arXiv:1808.01900v2 fatcat:rinp5iqn6ffopf3qhsjctno2km

Sequential Learning of Visual Tracking and Mapping Using Unsupervised Deep Neural Networks [article]

Youngji Kim, Ayoung Kim
2019 arXiv   pre-print
We proposed an end-to-end deep learning-based simultaneous localization and mapping (SLAM) system following conventional visual odometry (VO) pipelines.  ...  The experiment demonstrated that our method works well in tracking and mapping tasks and performs comparably with other learning-based VO approaches.  ...  DeepTAM [14] suggested mapping and tracking networks jointly working to accomplish localization and mapping within a key-frame.  ... 
arXiv:1902.09826v2 fatcat:zlvzcevgnba5tjauopkxxzomuy

TANDEM: Tracking and Dense Mapping in Real-time using Deep Multi-view Stereo [article]

Lukas Koestler, Nan Yang, Niclas Zeller, Daniel Cremers
2021 arXiv   pre-print
In this paper, we present TANDEM a real-time monocular tracking and dense mapping framework. For pose estimation, TANDEM performs photometric bundle adjustment based on a sliding window of keyframes.  ...  Our experimental results show that TANDEM outperforms other state-of-the-art traditional and learning-based monocular visual odometry (VO) methods in terms of camera tracking.  ...  Furthermore, TANDEM performs favourably in comparison to the tracking component of DeepTAM [23] that uses ground truth depth maps.  ... 
arXiv:2111.07418v1 fatcat:ht6kzjazvbhclkipoqb3ewj3zu

DeepV2D: Video to Depth with Differentiable Structure from Motion [article]

Zachary Teed, Jia Deng
2020 arXiv   pre-print
We propose DeepV2D, an end-to-end deep learning architecture for predicting depth from video.  ...  DeepV2D interleaves two stages: motion estimation and depth estimation. During inference, motion and depth estimation are alternated and converge to accurate depth.  ...  D ADDITIONAL TRACKING INFORMATION In We outperform DeepTAM and DVO on 12 of the 16 sequences and achieve a lower translational RMSE averaged over all sequences.  ... 
arXiv:1812.04605v4 fatcat:sgmbcd3tvjbcxitja6kc5jw2pm

DeepFactors: Real-Time Probabilistic Dense Monocular SLAM

Jan Czarnowski, Tristan Laidlow, Ronald Clark, Andrew J. Davison
2020 IEEE Robotics and Automation Letters  
This is achieved through the use of a learned compact depth map representation and reformulating three different types of errors: photometric, reprojection and geometric, which we make use of within standard  ...  Different approaches have been proposed that vary in scene geometry representation (sparse landmarks, dense maps), the consistency metric used for optimising the multi-view problem, and the use of learned  ...  A notable mention is DeepTAM [24] , which builds upon DTAM [7] by replacing both the TV-L1 optimisation and camera tracking with a deep convolutional neural network and achieves results outperforming  ... 
doi:10.1109/lra.2020.2965415 fatcat:43of2abopndmnavz3eywyvvplq

MonoRec: Semi-Supervised Dense Reconstruction in Dynamic Environments from a Single Moving Camera [article]

Felix Wimbauer, Nan Yang, Lukas von Stumberg, Niclas Zeller, Daniel Cremers
2021 arXiv   pre-print
In this paper, we propose MonoRec, a semi-supervised monocular dense reconstruction architecture that predicts depth maps from a single moving camera in dynamic environments.  ...  We carefully evaluate MonoRec on the KITTI dataset and show that it achieves state-of-the-art performance compared to both multi-view and single-view methods.  ...  Compared to DeepMVS and DeepTAM, MonoRec delivers depth maps with less artifacts and predicts the moving object masks in addition. D S t (x) , D S t (x) Dt(x) } > 1.5.  ... 
arXiv:2011.11814v3 fatcat:mtespeh55vdcdfbya6jrwtqjwi

DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Cameras [article]

Zachary Teed, Jia Deng
2022 arXiv   pre-print
We introduce DROID-SLAM, a new deep learning based SLAM system. DROID-SLAM consists of recurrent iterative updates of camera pose and pixelwise depth through a Dense Bundle Adjustment layer.  ...  DROID-SLAM is accurate, achieving large improvements over prior work, and robust, suffering from substantially fewer catastrophic failures.  ...  combines the strengths of both classical approaches and deep networks.  ... 
arXiv:2108.10869v2 fatcat:c3siuqdvercqfd63gqvatmoy3u

Beyond Tracking: Selecting Memory and Refining Poses for Deep Visual Odometry [article]

Fei Xue, Xin Wang, Shunkai Li, Qiuyuan Wang, Junqiu Wang, Hongbin Zha
2019 arXiv   pre-print
Most previous learning-based visual odometry (VO) methods take VO as a pure tracking problem.  ...  Experiments on the KITTI and TUM-RGBD benchmark datasets demonstrate that our method outperforms state-of-the-art learning-based methods by a large margin and produces competitive results against classic  ...  Acknowledgement The work is supported by the National Key Research and Development Program of China (2017YFB1002601) and National Natural Science Foundation of China (61632003, 61771026).  ... 
arXiv:1904.01892v2 fatcat:y6w24oqebfeu3hs7fotm22rwhm

Deep Visual Odometry with Adaptive Memory [article]

Fei Xue and Xin Wang and Junqiu Wang and Hongbin Zha
2020 arXiv   pre-print
We propose a novel deep visual odometry (VO) method that considers global information by selecting memory and refining poses.  ...  Specifically, our architecture consisting of Tracking, Remembering and Refining modules works beyond tracking.  ...  ACKNOWLEDGMENTS The work is supported by the National Key Research and Development Program of China (2017YFB1002601) and National Natural Science Foundation of China (61632003, 61771026).  ... 
arXiv:2008.01655v1 fatcat:oivwiy5oyrgntkry46awtwyvlq

Deep Probabilistic Feature-metric Tracking [article]

Binbin Xu, Andrew J. Davison, Stefan Leutenegger
2020 arXiv   pre-print
In this paper, we propose a new framework to learn a pixel-wise deep feature map and a deep feature-metric uncertainty map predicted by a Convolutional Neural Network (CNN), which together formulate a  ...  Experimental results demonstrate state-of-the-art performances on the TUM RGB-D dataset and the 3D rigid object tracking dataset.  ...  ACKNOWLEDGMENTS We wish to thank Shuaifeng Zhi, Jan Czarnowski and Tristan Laidlow for fruitful discussions.  ... 
arXiv:2008.13504v2 fatcat:eo4krefmofaw3pxum23pblp3ym

Accelerating the Convergence of Human-in-the-Loop Reinforcement Learning with Counterfactual Explanations [article]

Jakob Karalus, Felix Lindner
2021 arXiv   pre-print
For example, novice users could train service robots in new tasks naturally and interactively.  ...  Human-in-the-loop Reinforcement Learning (HRL) addresses this issue by combining human feedback and reinforcement learning (RL) techniques.  ...  However, this method can only be applied in specific cases due to Saliency Maps and Deep Neural Nets (solving vision problems with Convolution Neural Nets).  ... 
arXiv:2108.01358v1 fatcat:ufhrc7oy55cgrcirtdyhbfwske

Learning Meshes for Dense Visual SLAM

Michael Bloesch, Tristan Laidlow, Ronald Clark, Stefan Leutenegger, Andrew Davison
2019 2019 IEEE/CVF International Conference on Computer Vision (ICCV)  
This so-called factor graph encodes all information as mapping from free variables to residuals, the squared sum of which is minimised during inference.  ...  To allow for simple and fast usage, we propose a view-based formulation for which we predict the in-plane vertex coordinates directly from images and then employ the remaining vertex depth components as  ...  DeepTAM [39] uses separate tracking and mapping networks, the mapping network taking in a DTAM-style cost volume and then refining the prediction with several learning-based modules.  ... 
doi:10.1109/iccv.2019.00595 dblp:conf/iccv/BloeschLCLD19 fatcat:ixf64gexsbep3logwnnolxyx3m

RidgeSfM: Structure from Motion via Robust Pairwise Matching Under Depth Uncertainty [article]

Benjamin Graham, David Novotny
2020 arXiv   pre-print
We consider the problem of simultaneously estimating a dense depth map and camera pose for a large set of images of an indoor scene.  ...  To this end, we parametrize each depth map with a linear combination of a limited number of basis "depth-planes" predicted in a monocular fashion by a deep net.  ...  PTAM [22] was one of the first practical systems that allowed real-time tracking and mapping using a pair of reconstruction and tracking threads.  ... 
arXiv:2011.10359v1 fatcat:qlyuv4jvnvg7tkhk3uhledclmq

DF-VO: What Should Be Learnt for Visual Odometry? [article]

Huangying Zhan, Chamara Saroj Weerasekera, Jia-Wang Bian, Ravi Garg, Ian Reid
2021 arXiv   pre-print
Building on top of recent progress in computer vision, we design a simple yet robust VO system by integrating multi-view geometry and deep learning on Depth and optical Flow, namely DF-VO.  ...  Recent studies show that deep neural networks can learn scene depths and relative camera in a self-supervised manner without acquiring ground truth labels.  ...  Acknowledgment This work was supported by the UoA Scholarship to HZ, the ARC Laureate Fellowship FL130100102 to IR and the Australian Centre of Excellence for Robotic Vision CE140100016.  ... 
arXiv:2103.00933v1 fatcat:gs4bsysoozelfmv7mdntx3xiba
« Previous Showing results 1 — 15 out of 35 results