32,606 Hits in 4.9 sec

Deep Two-View Structure-from-Motion Revisited [article]

Jianyuan Wang, Yiran Zhong, Yuchao Dai, Stan Birchfield, Kaihao Zhang, Nikolai Smolyanskiy, Hongdong Li
2021 arXiv   pre-print
Two-view structure-from-motion (SfM) is the cornerstone of 3D reconstruction and visual SLAM.  ...  In contrast, we propose to revisit the problem of deep two-view SfM by leveraging the well-posedness of the classic pipeline.  ...  We would like to thank Shihao Jiang, Dylan Campbell, Charles Loop for helpful discussions and Ke Chen for providing field test images from NVIDIA AV cars.  ... 
arXiv:2104.00556v1 fatcat:tubxrltigvahrh7zlk3bnuckty

Degeneracy in Self-Calibration Revisited and a Deep Learning Solution for Uncalibrated SLAM [article]

Bingbing Zhuang, Quoc-Huy Tran, Pan Ji, Gim Hee Lee, Loong Fah Cheong, Manmohan Chandraker
2019 arXiv   pre-print
In this paper, we revisit the geometric approach to this problem, and provide a theoretical proof that explicitly shows the ambiguity between radial distortion and scene depth when two-view geometry is  ...  In view of such geometric degeneracy, we propose a learning approach that trains a convolutional neural network (CNN) on a large amount of synthetic data.  ...  III that estimating radial distortion from radially-distorted two-view geometry is degenerate when the camera undergoes ideal forward motion.  ... 
arXiv:1907.13185v1 fatcat:65esm7iluzdwdjg6yjcarpceuu

Visual Odometry Revisited: What Should Be Learnt? [article]

Huangying Zhan, Chamara Saroj Weerasekera, Jiawang Bian, Ian Reid
2020 arXiv   pre-print
Specifically, we train two convolutional neural networks (CNNs) for estimating single-view depths and two-view optical flows as intermediate outputs.  ...  More importantly, our system does not suffer from the scale-drift issue being aided by a scale consistent single-view depth CNN.  ...  Deep Predictions for Correspondence Suppose we have two deep models, M d and M f , for single-view depth estimation and two-view optical flow estimation respectively.  ... 
arXiv:1909.09803v4 fatcat:2yf6ozwtmbgj3bcete3ob4qgt4

End-to-End Multi-View Structure-from-Motion with Hypercorrelation Volumes [article]

Qiao Chen, Charalambos Poullis
2022 arXiv   pre-print
In this paper, we improve on the state-of-the-art two-view structure-from-motion(SfM) approach of [11] by incorporating 4D correlation volume for more accurate feature matching and reconstruction.  ...  Recently, deep learning techniques have been proposed to tackle this problem.  ...  In this paper, we extend a deep two-view structure-frommotion method to the more general task of multi-view reconstruction.  ... 
arXiv:2209.06926v1 fatcat:lruyyagfbzheblf5qs6kwln77q

Page 156 of The Geographical Journal Vol. 15, Issue 2 [page]

1900 The Geographical Journal  
vertical, the beds of apparent stratification being in planes perpendicular to the lines of pressure, quite in accordance with the accepted views of the method of the formatidn of glacier ice-structure  ...  for the whole period between May 21 and October 15, when it was revisited, the motion was 98^ feet, giving an average rate of just 8 inches in the 24 hours.  ... 

BA-Net: Dense Bundle Adjustment Network [article]

Chengzhou Tang, Ping Tan
2019 arXiv   pre-print
This paper introduces a network architecture to solve the structure-from-motion (SfM) problem via feature-metric bundle adjustment (BA), which explicitly enforces multi-view geometry constraints in the  ...  The whole system nicely combines domain knowledge (i.e. hard-coded multi-view geometry constraints) and deep learning (i.e. feature learning and basis depth maps learning) to address the challenging dense  ...  Instead, most existing deep learning approaches can only handle two views at a time, which is sub-optimal as known in structure-from-motion literature.  ... 
arXiv:1806.04807v3 fatcat:5v73f2buhnetjeucctrdt6jp34

Revisiting Video Saliency: A Large-scale Benchmark and a New Model [article]

Wenguan Wang and Jianbing Shen and Fang Guo and Ming-Ming Cheng and Ali Borji
2018 arXiv   pre-print
In this work, we contribute to video saliency research in two ways.  ...  First, we introduce a new benchmark for predicting human eye movements during dynamic scene free-viewing, which is long-time urged in this field.  ...  OM-CNN, Two-stream, SALICON, DVA, Shallow-Net, and Deep-Net are deep learning models, and others are classical saliency models.  ... 
arXiv:1801.07424v3 fatcat:kiutlpo76zbnrnyoemha5f25mu

Optical Synoptic Telescopes: New Science Frontiers

J. Anthony Tyson
2010 arXiv   pre-print
The history of astronomy has taught us repeatedly that there are surprises whenever we view the sky in a new way.  ...  This will be particularly true of discoveries emerging from a new generation of sky surveys.  ...  Joint baryon acoustic oscillation and weak gravitational lensing constraints on the comoving distance (triangles) and dark matter structure growth factor (circles) from a 20,000 deg 2 deep survey in six  ... 
arXiv:1009.2263v1 fatcat:b2f2hdntdzfptktucr5ve63tbu

DeepSFM: Structure From Motion Via Deep Bundle Adjustment [article]

Xingkui Wei, Yinda Zhang, Zhuwen Li, Yanwei Fu, Xiangyang Xue
2020 arXiv   pre-print
Structure from motion (SfM) is an essential computer vision problem which has not been well handled by deep learning.  ...  The explicit constraints on both depth (structure) and pose (motion), when combined with the learning components, bring the merit from both traditional BA and emerging deep learning technology.  ...  Deep Learning for Structure-from-Motion Deep neural networks have shown great success in stereo matching and Structure-from-Motion problems. [43,47,44,54] regress depth and camera pose directly in a supervised  ... 
arXiv:1912.09697v2 fatcat:4eo7c5m555f55hifys6eycsuvi

Large Synoptic Survey Telescope: From science drivers to reference design

Z. Ivezic, T. Axelrod, W.N. Brandt, D.L. Burke, C.F. Claver, A. Connolly, K.H. Cook, P. Gee, D.K. Gilmore, S.H. Jacoby, R.L. Jones, S.M. Kahn (+8 others)
2008 Serbian Astronomical Journal  
A visit time of 30 seconds (using two 15 second exposures to efficiently reject cosmic rays; = 77%). 3. A revisit time of 3 days on average per 10,000 deg 2 of sky, with two visits per night.  ...  We still lack robust answers to two basic questions about the Milky Way Galaxy: o What is the detailed structure and accretion history of the Milky Way?  ... 
doi:10.2298/saj0876001i fatcat:zrlcy6dq65avtla4br53rvgzmi

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.  ...  In this work, a) we propose a method to carefully sample high-quality correspondences from deep flows and recover accurate camera poses with a geometric module; b) we address the scale-drift issue by aligning  ...  Recently deep learning methods have made possible end-to-end learning of structure-and-motion from unlabelled videos.  ... 
arXiv:2103.00933v1 fatcat:gs4bsysoozelfmv7mdntx3xiba

Mapillary Street-Level Sequences: A Dataset for Lifelong Place Recognition

Frederik Warburg, Soren Hauberg, Manuel Lopez-Antequera, Pau Gargallo, Yubin Kuang, Javier Civera
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
It contains more than 1.6 million images curated from the Mapillary collaborative mapping platform.  ...  We contribute with Mapillary Street-Level Sequences (MSLS), a large dataset for urban and suburban place recognition from image sequences.  ...  It received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (grant agreement n o 757360), from the Spanish Government (PGC2018  ... 
doi:10.1109/cvpr42600.2020.00270 dblp:conf/cvpr/WarburgHLGKC20 fatcat:3zcatm2jurfhjhry6ym45j5i7i

One (sub-)acinus for all: Fate of inhaled aerosols in heterogeneous pulmonary acinar structures

Philipp Hofemeier, Kenishiro Koshiyama, Shigeo Wada, Josué Sznitman
2018 European Journal of Pharmaceutical Sciences  
With such tools at hand, we revisit whole-lung deposition predictions (i.e. ICRP) based on past 1D lung models.  ...  consolidate our mechanistic understanding of inhaled aerosol transport in the acinar depths but also continue to bridge the gap between "bottom-up" in silico models and regional deposition predictions from  ...  This article is based upon work from COST Action MP1404 SimInhale 'Simulation and pharmaceutical technologies for advanced patient-tailored inhaled medicines', supported by COST (European Cooperation in  ... 
doi:10.1016/j.ejps.2017.09.033 pmid:28954217 pmcid:PMC5810902 fatcat:jcp6sq6wrzc7tf3oikgp5xsdxu

Robust 3D street-view reconstruction using sky motion estimation

Taehee Lee
2009 2009 IEEE 12th International Conference on Computer Vision Workshops, ICCV Workshops  
We introduce a robust 3D reconstruction system that uses a combination of the structure-from-motion (SfM) filter and the bundle adjustment.  ...  In addition, we increase the robustness of the rotation estimation by estimating the motion of the sky from cylindrical panoramas of street views.  ...  Acknowledgment We would like to thank Google Research for providing the Street-View dataset for the research.  ... 
doi:10.1109/iccvw.2009.5457506 dblp:conf/iccvw/000209 fatcat:zpls45r76faafmrthb5lny3xzu

Ocean Measurements from Space in 2025

Anthony Freeman, Victor Zlotnicki, Tim Liu, Ben Holt, Ron Kwok, Simon Yueh, Jorge Vazquez, Dave Siegel, Gary Lagerloef
2010 Oceanography  
two platforms flying in formation, and gravity measurements from multiple platforms • Laser interferometry to improve the accuracy of gravity measurements from future GRACE-like missions • Wide-field-of-view  ...  from passive microwave and optical/infrared sensors, (3) fine-scale ice motion from SAR, ice freeboard with both radar and laser altimeters.  ... 
doi:10.5670/oceanog.2010.12 fatcat:vvdbyzp7jfc65odkgwtjgbvrne
« Previous Showing results 1 — 15 out of 32,606 results