Filters








2,048 Hits in 6.9 sec

End-to-end Learning of Cost-Volume Aggregation for Real-time Dense Stereo [article]

Andrey Kuzmin, Dmitry Mikushin, Victor Lempitsky
2016 arXiv   pre-print
We present a new deep learning-based approach for dense stereo matching.  ...  At the same time, our approach uses a deep convolutional network to predict the local parameters of cost volume aggregation process, which in this paper we implement using differentiable domain transform  ...  CONCLUSION We proposed a new method of computing dense stereo correspondences using convolutional neural network trained to aggregate the cost volume.  ... 
arXiv:1611.05689v1 fatcat:4y7sg7e2qvdxvkj4gx6elwyyf4

Multi-scale Iterative Residuals for Fast and Scalable Stereo Matching [article]

Kumail Raza, René Schuster, Didier Stricker
2021 arXiv   pre-print
Despite the remarkable progress of deep learning in stereo matching, there exists a gap in accuracy between real-time models and slower state-of-the-art models which are suitable for practical applications  ...  To reduce the computational cost of matching, we use multi-scale warped features to estimate disparity residuals and push the disparity search range in the cost volume to a minimum limit.  ...  ACKNOWLEDGMENTS This work was partially funded by the Federal Ministry of Education and Research Germany under the project DECODE (01IW21001).  ... 
arXiv:2110.12769v1 fatcat:j3qz62cul5fktpvhhjcw5u2wva

DeepPruner: Learning Efficient Stereo Matching via Differentiable PatchMatch [article]

Shivam Duggal, Shenlong Wang, Wei-Chiu Ma, Rui Hu, Raquel Urtasun
2019 arXiv   pre-print
Our goal is to significantly speed up the runtime of current state-of-the-art stereo algorithms to enable real-time inference.  ...  We then exploit this representation to learn which range to prune for each pixel.  ...  Finally, we describe how to aggregate the cost, refine the estimation, and perform end-toend learning. We refer the reader to Fig. 1 for an illustration of our approach.  ... 
arXiv:1909.05845v1 fatcat:qb5kigyqwfbfznyx57fmmbxogq

Fast Deep Stereo with 2D Convolutional Processing of Cost Signatures [article]

Kyle Yee, Ayan Chakrabarti
2019 arXiv   pre-print
This leads to longer running times to process each image pair, making them impractical for real-time use in robots and autonomous vehicles.  ...  these 'cost-signature' features to produce a dense disparity map.  ...  our approach to multi-view stereo-we believe the higher computational efficiency will make it possible to reason about correspondences across multiple cameras in real time, while bringing gains in depth  ... 
arXiv:1903.04939v1 fatcat:uxw2rzfgn5bcrgwxmc5lxbb3ue

StereoVoxelNet: Real-Time Obstacle Detection Based on Occupancy Voxels from a Stereo Camera Using Deep Neural Networks [article]

Hongyu Li, Zhengang Li, Neset Unver Akmandor, Huaizu Jiang, Yanzhi Wang, Taskin Padir
2022 arXiv   pre-print
constraints for real-time feedback.  ...  To counter the scarcity of high-quality real-world indoor stereo datasets, we collect a 1.36 hours stereo dataset with a Jackal robot which is used to fine-tune our model.  ...  The authors would like to thank Ying Wang, Nathaniel Hanson, Mingxi Jia, and Chenghao Wang for their help and discussions.  ... 
arXiv:2209.08459v1 fatcat:fmzk7ff3u5aoblzs5xdsuyknlm

MSDC-Net: Multi-Scale Dense and Contextual Networks for Automated Disparity Map for Stereo Matching [article]

Zhibo Rao and Mingyi He and Yuchao Dai and Zhidong Zhu and Bo Li and Renjie He
2019 arXiv   pre-print
To predict accurate disparity map, we propose a novel deep learning architecture for detectingthe disparity map from a rectified pair of stereo images, called MSDC-Net.  ...  The multi-scale residual 3D convolution module learns the different scale geometry context from the cost volume which aggregated by the multi-scale fusion 2D convolution module.  ...  Then we concatenate the different size features to obtain the aggregating features volume. Multi-scale feature fusion part fuses the aggregated features volume to form a cost volume.  ... 
arXiv:1904.12658v2 fatcat:3ohw3qmkvrbvthpcuis2biqe3m

Bidirectional Stereo Matching Network with Double Cost Volumes

Xiaogang Jia, Wei Chen, Zhengfa Liang
2021 IEEE Access  
Therefore, we propose BSDCNet, a real-time stereo matching network consisting of two main modules: Double Matching Cost Computation and Bidirectional Cost Aggregation Network.  ...  real-time and accurate stereo matching result.  ...  ACKNOWLEDGMENT The authors would like to all anonymous reviewers for their insightful review of and valuable comments on the manuscript, which helped to improve the quality of the paper.  ... 
doi:10.1109/access.2021.3050540 fatcat:gwor4mclrbci7lusww5lficxiy

On the Synergies between Machine Learning and Binocular Stereo for Depth Estimation from Images: a Survey [article]

Matteo Poggi, Fabio Tosi, Konstantinos Batsos, Philippos Mordohai, Stefano Mattoccia
2021 arXiv   pre-print
Stereo matching is one of the longest-standing problems in computer vision with close to 40 years of studies and research.  ...  Recently, the rise of machine learning and the rapid proliferation of deep learning enhanced stereo matching with new exciting trends and applications unthinkable until a few years ago.  ...  These measures are particularly appealing when the full cost volume is not available, e.g. when using off-the-shelf stereo or end-to-end models not having a cost volume at all. LFN [39] .  ... 
arXiv:2004.08566v2 fatcat:wcwfgzibo5evbkun3atpsz6kwm

Fast Deep Stereo with 2D Convolutional Processing of Cost Signatures

Kyle Yee, Ayan Chakrabarti
2020 2020 IEEE Winter Conference on Applications of Computer Vision (WACV)  
This leads to longer running times to process each image pair, making them impractical for real-time use in robots and autonomous vehicles.  ...  process these "cost-signature" features to produce a dense disparity map.  ...  IIS-1820693, including with an REU supplement for KY's participation in the project at Washington University.  ... 
doi:10.1109/wacv45572.2020.9093273 dblp:conf/wacv/YeeC20 fatcat:detjpbn33bee5hoeomp33aqlim

A novel stereo matching pipeline with robustness and unfixed disparity search range [article]

Jiazhi Liu, Feng Liu
2022 arXiv   pre-print
Stereo matching is an essential basis for various applications, but most stereo matching methods have poor generalization performance and require a fixed disparity search range.  ...  The new stereo matching pipeline have the following advantages: It 1) has better generalization performance than most of the current stereo matching methods; 2) relaxes the limitation of a fixed disparity  ...  After building a 3D cost volume, prior methods usually execute cost aggregation on the cost volume, so they need to store the volume at the whole inference process.  ... 
arXiv:2204.04865v2 fatcat:fv4ppjw6tbednoxaztdufhrma4

Bilateral Grid Learning for Stereo Matching Networks [article]

Bin Xu, Yuhua Xu, Xiaoli Yang, Wei Jia, Yulan Guo
2021 arXiv   pre-print
The slicing layer is parameter-free, which allows us to obtain a high quality cost volume of high resolution from a low-resolution cost volume under the guide of the learned guidance map efficiently.  ...  Real-time performance of stereo matching networks is important for many applications, such as automatic driving, robot navigation and augmented reality (AR).  ...  Given a pair of stereo images, the purpose of stereo matching is to establish the dense correspondences between the pixels in the left and right images.  ... 
arXiv:2101.01601v2 fatcat:7t5chd2yanf7tedvhqaaxn4e7u

Bayesian Learning for Disparity Map Refinement for Semi-Dense Active Stereo Vision [article]

Laurent Valentin Jospin, Hamid Laga, Farid Boussaid, Mohammed Bennamoun
2022 arXiv   pre-print
In this paper, we propose a new learning strategy to train neural networks to estimate high-quality subpixel disparity maps for semi-dense active stereo vision.  ...  Active vision systems enable more accurate estimations of dense disparity compared to passive stereo.  ...  This paper contribution relates to the last step (see Fig. 3 ). end-to-end deep learning approaches for stereo matching, the cost volume is generally interpolated on the fly via the use of the softargmax  ... 
arXiv:2209.05082v1 fatcat:j4xmsmhfovewdpiwb3zsipl24i

A Convolutional Attention Residual Network for Stereo Matching

Guangyi Huang, Yongyi Gong, Qingzhen Xu, Kanoksak Wattanachote, Kun Zeng, Xiaonan Luo
2020 IEEE Access  
Moreover, we proposed a dimension-extended 3D-CBAM, which is connected to 3DCNN for cost aggregation.  ...  In this paper, we proposed a lightweight network, convolution attention residual network (CAR-Net), which can balance the real-time matching and matching accuracy for stereo matching.  ...  forward an urgent demand for the use of end-to-end deep learning network to solve stereo matching problems.  ... 
doi:10.1109/access.2020.2980243 fatcat:qofjoaxl45ed5nckwyyfu43xt4

Real-Time Semantic Segmentation-Based Stereo Reconstruction

Vlad-Cristian Miclea, Sergiu Nedevschi
2019 IEEE transactions on intelligent transportation systems (Print)  
In this paper, we propose a novel semantic segmentation-based stereo reconstruction method that can keep up with the accuracy of the state-of-the art approaches while running in real time.  ...  For the cost computation and optimization steps, we propose new genetic algorithms that can incrementally adjust the parameters for better solutions.  ...  INTRODUCTION D UE to the fast evolution of intelligent vehicles, real-time depth perception has become a major area of interest.  ... 
doi:10.1109/tits.2019.2913883 fatcat:jl2t5lit2nfmnfilv6lrbwz4pi

Matching-space Stereo Networks for Cross-domain Generalization [article]

Changjiang Cai, Matteo Poggi, Stefano Mattoccia, Philippos Mordohai
2020 arXiv   pre-print
End-to-end deep networks represent the state of the art for stereo matching.  ...  By replacing learning-based feature extraction from image RGB values with matching functions and confidence measures from conventional wisdom, we move the learning process from the color space to the Matching  ...  We gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan Xp GPU used for this research.  ... 
arXiv:2010.07347v1 fatcat:t5x6abwc2nhhha4owii32rispi
« Previous Showing results 1 — 15 out of 2,048 results