Search all papers or click the papers filter menu above for a selective search.
1.
True Price Jisan Mahmud, Akash Bapat; Frahm, Jan-Michael
Boundary-aware 3D Building Reconstruction from a Single Overhead Image: Supplementary Journal Article
In: 2020.
@article{cs20-001,
title = {Boundary-aware 3D Building Reconstruction from a Single Overhead Image: Supplementary},
author = {Jisan Mahmud, True Price, Akash Bapat and Jan-Michael Frahm},
url = {https://techreports.cs.unc.edu/papers/20-001.pdf},
year = {2020},
date = {2020-03-24},
urldate = {1970-01-01},
abstract = {In this supplementary material, we present additional details about the network, optimization, and building overlap refinement strategies of our method. We also present additional qualitative results and ablative analysis.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In this supplementary material, we present additional details about the network, optimization, and building overlap refinement strategies of our method. We also present additional qualitative results and ablative analysis.
2.
Yan, Junhua; Kaur, Jasleen
Feature Selection for Website Fingerprinting Technical Report
no. 18-001, 2018.
@techreport{cs18-001,
title = {Feature Selection for Website Fingerprinting},
author = {Junhua Yan and Jasleen Kaur},
url = {https://techreports.cs.unc.edu/papers/18-001.pdf},
year = {2018},
date = {2018-06-05},
urldate = {1970-01-01},
number = {18-001},
abstract = {Website fingerprinting based on TCP/IP headers is of significant relevance to several Internet entities. Prior work has focused only on a limited set of features, and does not help understand the extents of fingerprint-ability. We address this by conducting an exhaustive feature analysis within eight different communication scenarios. Our analysis helps reveal several previously-unknown features in several scenarios, that can be used to fingerprint websites with much higher accuracy than previously demonstrated. This work helps the community better understand the extents of learnability
(and vulnerability) from TCP/IP headers.
Keywords: website fingerprinting, feature selection},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
Website fingerprinting based on TCP/IP headers is of significant relevance to several Internet entities. Prior work has focused only on a limited set of features, and does not help understand the extents of fingerprint-ability. We address this by conducting an exhaustive feature analysis within eight different communication scenarios. Our analysis helps reveal several previously-unknown features in several scenarios, that can be used to fingerprint websites with much higher accuracy than previously demonstrated. This work helps the community better understand the extents of learnability
(and vulnerability) from TCP/IP headers.
Keywords: website fingerprinting, feature selection
(and vulnerability) from TCP/IP headers.
Keywords: website fingerprinting, feature selection
3.
Wang, Chun-Kun; Xu, Hao
A Survey of Reasoning Methods for Concurrent Systems Technical Report
no. 17-005, 2017.
@techreport{cs17-005,
title = {A Survey of Reasoning Methods for Concurrent Systems},
author = {Chun-Kun Wang and Hao Xu},
url = {https://techreports.cs.unc.edu/papers/17-005.pdf},
year = {2017},
date = {2017-12-14},
urldate = {1970-01-01},
number = {17-005},
abstract = {The theory of parallel and distributed systems in computer science has been developing for decades. Multiple methods have emerged for the purpose of providing flexibility, expressiveness and theories for concurrent processes. This paper provides a comparative introduction to the history of concurrent systems reasoning and a decision tree to guide the choice of a reasoning method accordingly. Our account on these matters is incomplete.
Index Terms--Concurrent systems, Formal methods, Program logic, Process calculi.},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
The theory of parallel and distributed systems in computer science has been developing for decades. Multiple methods have emerged for the purpose of providing flexibility, expressiveness and theories for concurrent processes. This paper provides a comparative introduction to the history of concurrent systems reasoning and a decision tree to guide the choice of a reasoning method accordingly. Our account on these matters is incomplete.
Index Terms--Concurrent systems, Formal methods, Program logic, Process calculi.
Index Terms--Concurrent systems, Formal methods, Program logic, Process calculi.
4.
Islam, Bashima; Islam, MD Tamzeed; Nirjon, Shahriar
Feasibility of LoRa for Indoor Localization Technical Report
no. 17-004, 2017.
@techreport{cs17-004,
title = {Feasibility of LoRa for Indoor Localization},
author = {Bashima Islam and MD Tamzeed Islam and Shahriar Nirjon},
url = {https://techreports.cs.unc.edu/papers/17-004.pdf},
year = {2017},
date = {2017-12-04},
urldate = {1970-01-01},
number = {17-004},
abstract = {Feasibility of LoRa for Indoor Localization},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
Feasibility of LoRa for Indoor Localization
5.
Islam, Tamzeed; Islam, Bashima; Nirjon, Shahriar
Duty-Cycle-Aware Real-Time Scheduling of Wireless Links in Low Power WANs Technical Report
no. 17-003, 2017.
@techreport{cs17-003,
title = {Duty-Cycle-Aware Real-Time Scheduling of Wireless Links in Low Power WANs},
author = {Tamzeed Islam and Bashima Islam and Shahriar Nirjon },
url = {https://techreports.cs.unc.edu/papers/17-003.pdf},
year = {2017},
date = {2017-11-22},
urldate = {1970-01-01},
number = {17-003},
abstract = {Low Power Wide Area Networks (LPWANs) are anexcellent fit to city-scale IoT applications—offering a long range,
long battery life of several years, and a data rate of 25-50kbps which is sufficient to carry IoT traffic. However, a practical limitation in realizing a LPWAN-based real-time wireless network is the duty-cycle limit that is imposed on the sub-1GHz band by the FCC. In this paper, we overcome this challenge by proposing the first duty-cycle-aware wireless link scheduling algorithm for real-time LPWANs that considers the urgency of the packets as well as the availability of the channels. The proposed algorithm is implemented in a five-node city-wide test-bed in multiple rea lworld scenarios. Simulation results are provided to quantify its performance under different settings (e.g. larger networks, variety of workloads, and multiple baselines). In both real-world deployments and simulations, the proposed scheduling algorithm has outperformed all the baselines in terms of link schedulability, deadline misses, and buffer size.},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
Low Power Wide Area Networks (LPWANs) are anexcellent fit to city-scale IoT applications—offering a long range,
long battery life of several years, and a data rate of 25-50kbps which is sufficient to carry IoT traffic. However, a practical limitation in realizing a LPWAN-based real-time wireless network is the duty-cycle limit that is imposed on the sub-1GHz band by the FCC. In this paper, we overcome this challenge by proposing the first duty-cycle-aware wireless link scheduling algorithm for real-time LPWANs that considers the urgency of the packets as well as the availability of the channels. The proposed algorithm is implemented in a five-node city-wide test-bed in multiple rea lworld scenarios. Simulation results are provided to quantify its performance under different settings (e.g. larger networks, variety of workloads, and multiple baselines). In both real-world deployments and simulations, the proposed scheduling algorithm has outperformed all the baselines in terms of link schedulability, deadline misses, and buffer size.
long battery life of several years, and a data rate of 25-50kbps which is sufficient to carry IoT traffic. However, a practical limitation in realizing a LPWAN-based real-time wireless network is the duty-cycle limit that is imposed on the sub-1GHz band by the FCC. In this paper, we overcome this challenge by proposing the first duty-cycle-aware wireless link scheduling algorithm for real-time LPWANs that considers the urgency of the packets as well as the availability of the channels. The proposed algorithm is implemented in a five-node city-wide test-bed in multiple rea lworld scenarios. Simulation results are provided to quantify its performance under different settings (e.g. larger networks, variety of workloads, and multiple baselines). In both real-world deployments and simulations, the proposed scheduling algorithm has outperformed all the baselines in terms of link schedulability, deadline misses, and buffer size.
6.
Otterness, Nathan; Yang, Ming; Amert, Tanya; Bakita, Joshua; Anderson, James H.; Smith, F. Donelson
Implicit GPU Synchronization: A Barrier to Real-Time CUDA Workloads Technical Report
no. 17-002, 2017.
@techreport{cs17-002,
title = {Implicit GPU Synchronization: A Barrier to Real-Time CUDA Workloads},
author = {Nathan Otterness and Ming Yang and Tanya Amert and Joshua Bakita and James H. Anderson and F. Donelson Smith},
year = {2017},
date = {2017-10-02},
urldate = {1970-01-01},
number = {17-002},
abstract = {NVIDIA’s CUDA API has enabled GPUs to
be used as computing accelerators across a wide range of
applications. This has resulted in performance gains in many
application domains, but the underlying GPU hardware and
software is subject to many non-obvious pitfalls. This is
particularly problematic for safety-critical systems, where
worst-case behaviors must be taken into account. While such
behaviors were not a key concern for earlier CUDA users, the
advent of autonomous vehicles has taken CUDA programs out
of the sole domain of computer-vision and machine-learning
experts and into safety-critical processing pipelines. Certification
is necessary in this new domain, which is problematic
because GPU software may have been developed without
any regard for worst-case behaviors. One potential pitfall for
real-time CUDA usage is significant blocking due to implicit
synchronization, where all pending GPU work must complete
before new work can begin. Due to its implicit nature and
the lack of specifics in official documentation, developers
of GPU software may not be aware of the effects of such
synchronization, much less the full extent of its worst-case
impact. This paper focuses on the causes and implications of
implicit GPU synchronization, its prominence in state-of-theart
compute-vision libraries, and approaches for ameliorating
its negative consequence in real-time analysis.},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
NVIDIA’s CUDA API has enabled GPUs to
be used as computing accelerators across a wide range of
applications. This has resulted in performance gains in many
application domains, but the underlying GPU hardware and
software is subject to many non-obvious pitfalls. This is
particularly problematic for safety-critical systems, where
worst-case behaviors must be taken into account. While such
behaviors were not a key concern for earlier CUDA users, the
advent of autonomous vehicles has taken CUDA programs out
of the sole domain of computer-vision and machine-learning
experts and into safety-critical processing pipelines. Certification
is necessary in this new domain, which is problematic
because GPU software may have been developed without
any regard for worst-case behaviors. One potential pitfall for
real-time CUDA usage is significant blocking due to implicit
synchronization, where all pending GPU work must complete
before new work can begin. Due to its implicit nature and
the lack of specifics in official documentation, developers
of GPU software may not be aware of the effects of such
synchronization, much less the full extent of its worst-case
impact. This paper focuses on the causes and implications of
implicit GPU synchronization, its prominence in state-of-theart
compute-vision libraries, and approaches for ameliorating
its negative consequence in real-time analysis.
be used as computing accelerators across a wide range of
applications. This has resulted in performance gains in many
application domains, but the underlying GPU hardware and
software is subject to many non-obvious pitfalls. This is
particularly problematic for safety-critical systems, where
worst-case behaviors must be taken into account. While such
behaviors were not a key concern for earlier CUDA users, the
advent of autonomous vehicles has taken CUDA programs out
of the sole domain of computer-vision and machine-learning
experts and into safety-critical processing pipelines. Certification
is necessary in this new domain, which is problematic
because GPU software may have been developed without
any regard for worst-case behaviors. One potential pitfall for
real-time CUDA usage is significant blocking due to implicit
synchronization, where all pending GPU work must complete
before new work can begin. Due to its implicit nature and
the lack of specifics in official documentation, developers
of GPU software may not be aware of the effects of such
synchronization, much less the full extent of its worst-case
impact. This paper focuses on the causes and implications of
implicit GPU synchronization, its prominence in state-of-theart
compute-vision libraries, and approaches for ameliorating
its negative consequence in real-time analysis.
7.
Yin, Qianwen; Kaur, Jasleen; Smith, Don
TCP Rapid: From Theory to Practice Technical Report
no. 17-001, 2017.
@techreport{cs17-001,
title = {TCP Rapid: From Theory to Practice},
author = {Qianwen Yin and Jasleen Kaur and Don Smith},
url = {https://techreports.cs.unc.edu/papers/17-001.pdf},
year = {2017},
date = {2017-01-12},
urldate = {1970-01-01},
number = {17-001},
abstract = { Delay and rate-based alternatives to TCP congestion-control have been around for nearly three decades and have seen a recent surge in interest. However, such designs have faced significant resistance in being deployed on a wide-scale across the Internet--his has been mostly due to serious concerns about noise in delay measurements, pacing inter-packet gaps, and/or required changes to the standard TCP stack/headers. With the advent of high-speed networking, some of these concerns become even more significant.
In this paper, we consider Rapid, a recent proposal for ultra-high speed congestion control, which perhaps stretches each of these challenges to the greatest extent. Rapid adopts a framework of continuous fine-scale bandwidth probing, which requires a potentially different and finely-controlled gap for every packet, high-precision timestamping of received packets, and reliance on fine-scale changes in inter-packet gaps. While simulation-based evaluations of Rapid show that it has outstanding performance gains along several important dimensions, these will not translate to the real-world unless the above challenges are addressed.
We design a Linux implementation of Rapid after carefully considering each of these challenges. Our evaluations on a 10Gbps test bed confirm that the implementation can indeed achieve the claimed performance gains, and that it would not have been possible unless each of the above challenges was addressed.},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
Delay and rate-based alternatives to TCP congestion-control have been around for nearly three decades and have seen a recent surge in interest. However, such designs have faced significant resistance in being deployed on a wide-scale across the Internet--his has been mostly due to serious concerns about noise in delay measurements, pacing inter-packet gaps, and/or required changes to the standard TCP stack/headers. With the advent of high-speed networking, some of these concerns become even more significant.
In this paper, we consider Rapid, a recent proposal for ultra-high speed congestion control, which perhaps stretches each of these challenges to the greatest extent. Rapid adopts a framework of continuous fine-scale bandwidth probing, which requires a potentially different and finely-controlled gap for every packet, high-precision timestamping of received packets, and reliance on fine-scale changes in inter-packet gaps. While simulation-based evaluations of Rapid show that it has outstanding performance gains along several important dimensions, these will not translate to the real-world unless the above challenges are addressed.
We design a Linux implementation of Rapid after carefully considering each of these challenges. Our evaluations on a 10Gbps test bed confirm that the implementation can indeed achieve the claimed performance gains, and that it would not have been possible unless each of the above challenges was addressed.
In this paper, we consider Rapid, a recent proposal for ultra-high speed congestion control, which perhaps stretches each of these challenges to the greatest extent. Rapid adopts a framework of continuous fine-scale bandwidth probing, which requires a potentially different and finely-controlled gap for every packet, high-precision timestamping of received packets, and reliance on fine-scale changes in inter-packet gaps. While simulation-based evaluations of Rapid show that it has outstanding performance gains along several important dimensions, these will not translate to the real-world unless the above challenges are addressed.
We design a Linux implementation of Rapid after carefully considering each of these challenges. Our evaluations on a 10Gbps test bed confirm that the implementation can indeed achieve the claimed performance gains, and that it would not have been possible unless each of the above challenges was addressed.
8.
Newton, Ben; Aikat, Jay; Jeffay, Kevin
Geographic Routing in Large-Scale Highly-Dynamic Mobile Ad-Hoc Networks Technical Report
no. 16-002, 2016.
@techreport{cs16-002,
title = {Geographic Routing in Large-Scale Highly-Dynamic Mobile Ad-Hoc Networks},
author = {Ben Newton and Jay Aikat and Kevin Jeffay},
url = {https://techreports.cs.unc.edu/papers/16-002.pdf},
year = {2016},
date = {2016-06-28},
urldate = {1970-01-01},
number = {16-002},
abstract = {In the near future extremely large scale mobile ad-hoc networks of thousands or tens of thousands of mobile node swill be physically feasible and desirable for a host of applications.However, routing within these networks is challenging, especially at high data rates and when node movement is highly-dynamic.In this work we present Topology Aware Geographic Routing(TAG), a position-based routing protocol which strategically uses local topology information (when available) to make better local forwarding decisions, decreasing the number of hops required to deliver a packet when compared with other geographic routing protocols. In addition TAG is able to reliably deliver packets even in topologies which violate the often used but unrealistic unit disk graph and quasi-static assumptions. We present empirical results from a variety of simulations, illustrating how TAG outperforms GOAFR+, GFG, and OLSR in both theoretical environments and in a simulated, real-world, continental-scale airborne network.},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
In the near future extremely large scale mobile ad-hoc networks of thousands or tens of thousands of mobile node swill be physically feasible and desirable for a host of applications.However, routing within these networks is challenging, especially at high data rates and when node movement is highly-dynamic.In this work we present Topology Aware Geographic Routing(TAG), a position-based routing protocol which strategically uses local topology information (when available) to make better local forwarding decisions, decreasing the number of hops required to deliver a packet when compared with other geographic routing protocols. In addition TAG is able to reliably deliver packets even in topologies which violate the often used but unrealistic unit disk graph and quasi-static assumptions. We present empirical results from a variety of simulations, illustrating how TAG outperforms GOAFR+, GFG, and OLSR in both theoretical environments and in a simulated, real-world, continental-scale airborne network.
9.
Rosenman, Julian; Zhao, Qingyu; Price, True; Niethammer, Marc; Alterovitz, Ron; Frahm, Jan-Michael; Chera, Bhishamjit; Pizer, Stephen
Registration Of Nasopharyngoscopic Video With Treatment Planning CT Scans Technical Report
no. 16-001, 2016.
@techreport{cs16-001,
title = {Registration Of Nasopharyngoscopic Video With Treatment Planning CT Scans},
author = {Julian Rosenman and Qingyu Zhao and True Price and Marc Niethammer and Ron Alterovitz and Jan-Michael Frahm and Bhishamjit Chera and Stephen Pizer},
url = {https://techreports.cs.unc.edu/papers/16-001.pdf},
year = {2016},
date = {2016-05-25},
urldate = {1970-01-01},
number = {16-001},
abstract = {Purpose: We present a method that allows one to generate data from nasopharyngoscopy videos into a form suitable for use in radiation treatment planning, as well as for review of the data itself as a 3D object.
Methods and Materials: The accurate transfer of data from endoscopic video frames to the planning CT requires an explicit endoscopic video/CT registration. Our approach to accomplishing this goal is to first reconstruct an accurate 3D model of the anatomy from the 2D video data as seen on multiple video frames. That reconstruction, which we call an endoscopogram, can then be deformably registered to the CT scan.
To generate the endoscopogram we first compute a sparse 3D point cloud by tracking corresponding feature points from one video frame to another using standard structure-from-motion (SfM) techniques adapted to the pharyngeal region. We then use a second technology, shape-from-shading (SFS), to produce high-quality images from each video frame. Uncorrected, these SFS images contain dense information and in many places accurate local curvature, but they lack global spatial fidelity. Using our newly developed techniques, we can correct the SFS images using global information from the sparse SFS-generated point cloud as a constraint. We then register several of these corrected SFS images to each other to generate a 3D image that displays the entire surface area seen in the whole video.
Results: We have successfully computed a spatially accurate endoscopogram (shown below) and have registered it with the planning CT scan with an average accuracy of about 3-5 mms.
Conclusions: The technologies developed by our working group will allow accurate transfer of data from endoscopy images directly onto the planning CT scan as well as review of the video as a 3D object.
},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
Purpose: We present a method that allows one to generate data from nasopharyngoscopy videos into a form suitable for use in radiation treatment planning, as well as for review of the data itself as a 3D object.
Methods and Materials: The accurate transfer of data from endoscopic video frames to the planning CT requires an explicit endoscopic video/CT registration. Our approach to accomplishing this goal is to first reconstruct an accurate 3D model of the anatomy from the 2D video data as seen on multiple video frames. That reconstruction, which we call an endoscopogram, can then be deformably registered to the CT scan.
To generate the endoscopogram we first compute a sparse 3D point cloud by tracking corresponding feature points from one video frame to another using standard structure-from-motion (SfM) techniques adapted to the pharyngeal region. We then use a second technology, shape-from-shading (SFS), to produce high-quality images from each video frame. Uncorrected, these SFS images contain dense information and in many places accurate local curvature, but they lack global spatial fidelity. Using our newly developed techniques, we can correct the SFS images using global information from the sparse SFS-generated point cloud as a constraint. We then register several of these corrected SFS images to each other to generate a 3D image that displays the entire surface area seen in the whole video.
Results: We have successfully computed a spatially accurate endoscopogram (shown below) and have registered it with the planning CT scan with an average accuracy of about 3-5 mms.
Conclusions: The technologies developed by our working group will allow accurate transfer of data from endoscopy images directly onto the planning CT scan as well as review of the video as a 3D object.
Methods and Materials: The accurate transfer of data from endoscopic video frames to the planning CT requires an explicit endoscopic video/CT registration. Our approach to accomplishing this goal is to first reconstruct an accurate 3D model of the anatomy from the 2D video data as seen on multiple video frames. That reconstruction, which we call an endoscopogram, can then be deformably registered to the CT scan.
To generate the endoscopogram we first compute a sparse 3D point cloud by tracking corresponding feature points from one video frame to another using standard structure-from-motion (SfM) techniques adapted to the pharyngeal region. We then use a second technology, shape-from-shading (SFS), to produce high-quality images from each video frame. Uncorrected, these SFS images contain dense information and in many places accurate local curvature, but they lack global spatial fidelity. Using our newly developed techniques, we can correct the SFS images using global information from the sparse SFS-generated point cloud as a constraint. We then register several of these corrected SFS images to each other to generate a 3D image that displays the entire surface area seen in the whole video.
Results: We have successfully computed a spatially accurate endoscopogram (shown below) and have registered it with the planning CT scan with an average accuracy of about 3-5 mms.
Conclusions: The technologies developed by our working group will allow accurate transfer of data from endoscopy images directly onto the planning CT scan as well as review of the video as a 3D object.
10.
Cheng, Wei
Toward Robust Group-wise Eqtl Mapping Via Integrating Multi-domain Heterogeneous Data Technical Report
no. 15-002, 2015.
@techreport{cs-15-002,
title = {Toward Robust Group-wise Eqtl Mapping Via Integrating Multi-domain Heterogeneous Data},
author = {Wei Cheng},
url = {https://techreports.cs.unc.edu/papers/14-005.pdf},
year = {2015},
date = {2015-12-01},
urldate = {1970-01-01},
number = {15-002},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
11.
Heinly, Jared
Toward Efficient and Robust Large-Scale Structure-from-Motion Systems Technical Report
no. 15-001, 2015.
@techreport{cs15-001,
title = {Toward Efficient and Robust Large-Scale Structure-from-Motion Systems},
author = {Jared Heinly},
url = {https://techreports.cs.unc.edu/papers/15-001.pdf},
year = {2015},
date = {2015-12-01},
urldate = {1970-01-01},
number = {15-001},
abstract = {The ever-increasing number of images that are uploaded and shared on the Internet has recently been leveraged by computer vision researchers to extract 3D information about the content seen in these images. One key mechanism to extract this information is structure-from-motion, which is the process of recovering the 3D geometry (structure) of a scene via a set of images from different viewpoints (camera motion). However, when dealing with crowdsourced datasets comprised of tens or hundreds of millions of images, the magnitude and diversity of the imagery poses challenges such as robustness, scalability, completeness, and correctness for existing structure-from-motion systems. This dissertation focuses on these challenges and demonstrates practical methods to address the problems of data association and verification within structure-from-motion systems.
Data association within structure-from-motion systems consists of the discovery of pairwise image overlap within the input dataset. In order to perform this discovery, previous systems assumed that information about every image in the input dataset could be stored in memory, which is prohibitive for large-scale photo collections. To address this issue, we propose a novel streaming-based framework for the discovery of related sets of images, and demonstrate our approach on a crowdsourced dataset containing 100 million images from all around the world. Results illustrate that our streaming-based approach does not compromise model completeness, but achieves unprecedented levels of efficiency and scalability.
The verification of individual data associations is difficult to perform during the process of structure-from-motion, as standard methods have limited scope when determining image overlap. Therefore, it is possible for erroneous associations to form, especially when there are symmetric, repetitive, or duplicate structures which can be incorrectly associated with each other. The consequences of these errors are incorrectly placed cameras and scene geometry within the 3D reconstruction. We present two methods that can detect these local inconsistencies and successfully resolve them into a globally consistent 3D model. In our evaluation, we show that our techniques are efficient, are robust to a variety of scenes, and outperform existing approaches.},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
The ever-increasing number of images that are uploaded and shared on the Internet has recently been leveraged by computer vision researchers to extract 3D information about the content seen in these images. One key mechanism to extract this information is structure-from-motion, which is the process of recovering the 3D geometry (structure) of a scene via a set of images from different viewpoints (camera motion). However, when dealing with crowdsourced datasets comprised of tens or hundreds of millions of images, the magnitude and diversity of the imagery poses challenges such as robustness, scalability, completeness, and correctness for existing structure-from-motion systems. This dissertation focuses on these challenges and demonstrates practical methods to address the problems of data association and verification within structure-from-motion systems.
Data association within structure-from-motion systems consists of the discovery of pairwise image overlap within the input dataset. In order to perform this discovery, previous systems assumed that information about every image in the input dataset could be stored in memory, which is prohibitive for large-scale photo collections. To address this issue, we propose a novel streaming-based framework for the discovery of related sets of images, and demonstrate our approach on a crowdsourced dataset containing 100 million images from all around the world. Results illustrate that our streaming-based approach does not compromise model completeness, but achieves unprecedented levels of efficiency and scalability.
The verification of individual data associations is difficult to perform during the process of structure-from-motion, as standard methods have limited scope when determining image overlap. Therefore, it is possible for erroneous associations to form, especially when there are symmetric, repetitive, or duplicate structures which can be incorrectly associated with each other. The consequences of these errors are incorrectly placed cameras and scene geometry within the 3D reconstruction. We present two methods that can detect these local inconsistencies and successfully resolve them into a globally consistent 3D model. In our evaluation, we show that our techniques are efficient, are robust to a variety of scenes, and outperform existing approaches.
Data association within structure-from-motion systems consists of the discovery of pairwise image overlap within the input dataset. In order to perform this discovery, previous systems assumed that information about every image in the input dataset could be stored in memory, which is prohibitive for large-scale photo collections. To address this issue, we propose a novel streaming-based framework for the discovery of related sets of images, and demonstrate our approach on a crowdsourced dataset containing 100 million images from all around the world. Results illustrate that our streaming-based approach does not compromise model completeness, but achieves unprecedented levels of efficiency and scalability.
The verification of individual data associations is difficult to perform during the process of structure-from-motion, as standard methods have limited scope when determining image overlap. Therefore, it is possible for erroneous associations to form, especially when there are symmetric, repetitive, or duplicate structures which can be incorrectly associated with each other. The consequences of these errors are incorrectly placed cameras and scene geometry within the 3D reconstruction. We present two methods that can detect these local inconsistencies and successfully resolve them into a globally consistent 3D model. In our evaluation, we show that our techniques are efficient, are robust to a variety of scenes, and outperform existing approaches.
12.
Bethea, Darrell; Reiter, Michael K; Qian, Feng; Xu, Qiang; Mao, Z Morley
Strong Consistency at Global Scale Technical Report
no. 14-004, 2014.
@techreport{CS14-004,
title = {Strong Consistency at Global Scale},
author = {Darrell Bethea and Michael K Reiter and Feng Qian and Qiang Xu and Z Morley Mao},
url = {https://techreports.cs.unc.edu/papers/14-004.pdf},
year = {2014},
date = {2014-12-19},
urldate = {1970-01-01},
number = {14-004},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
13.
Taylor, Micah; Tsingos, Nicolas; Manocha, Dinesh
Rendering environmental voice reverberation for large-scale distributed virtual worlds Technical Report
no. 14-005, 2014.
@techreport{CS14-005,
title = {Rendering environmental voice reverberation for large-scale distributed virtual worlds},
author = {Micah Taylor and Nicolas Tsingos and Dinesh Manocha},
url = {https://techreports.cs.unc.edu/papers/14-005.pdf},
year = {2014},
date = {2014-12-19},
urldate = {1970-01-01},
number = {14-005},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
14.
Lovewell, Rebecca; Yin, Qianwen; Zhang, Tao; Kaur, Jasleen; Smith, Frank
The packet-scale congestion control paradigm Technical Report
no. 14-006, 2014.
@techreport{CS14-006,
title = {The packet-scale congestion control paradigm},
author = {Rebecca Lovewell and Qianwen Yin and Tao Zhang and Jasleen Kaur and Frank Smith},
year = {2014},
date = {2014-07-15},
urldate = {1970-01-01},
number = {14-006},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
15.
Erickson, Jeremy; Anderson, James
Dissipation Bounds Technical Report
no. 14-001, 2014.
@techreport{CS14-001,
title = {Dissipation Bounds},
author = {Jeremy Erickson and James Anderson},
url = {https://techreports.cs.unc.edu/papers/14-001.pdf},
year = {2014},
date = {2014-05-06},
urldate = {1970-01-01},
number = {14-001},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
16.
Chou, Chen-Rui
Regression Learning for 2D/3D Image Registration Technical Report
no. 13-004, 2013.
@techreport{CS13-004,
title = {Regression Learning for 2D/3D Image Registration},
author = {Chen-Rui Chou},
url = {https://techreports.cs.unc.edu/papers/13-004.pdf},
year = {2013},
date = {2013-12-15},
urldate = {1970-01-01},
number = {13-004},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
17.
Libonati, Alana; Kapadia, Apu; Reiter, Michael K.
Social Security Technical Report
no. 13-003, 2013.
@techreport{CS13-003,
title = {Social Security},
author = {Alana Libonati and Apu Kapadia and Michael K. Reiter},
url = {https://techreports.cs.unc.edu/papers/13-003.pdf},
year = {2013},
date = {2013-10-08},
urldate = {1970-01-01},
number = {13-003},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
18.
Kohli, Luv
Redirected Touching Technical Report
no. 13-002, 2013.
@techreport{CS13-002,
title = {Redirected Touching},
author = {Luv Kohli},
url = {https://techreports.cs.unc.edu/papers/13-002.pdf},
year = {2013},
date = {2013-05-30},
urldate = {1970-01-01},
number = {13-002},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
19.
Bethea, Darrell; Reiter, Michael K.; Xu, Feng Qian; Mao, Z. Morley
WACCO and LOKO Technical Report
no. 12-005, 2013.
@techreport{CS12-005,
title = {WACCO and LOKO},
author = {Darrell Bethea and Michael K. Reiter and Feng Qian Xu and Z. Morley Mao},
url = {https://techreports.cs.unc.edu/papers/12-005R.pdf},
year = {2013},
date = {2013-01-08},
urldate = {1970-01-01},
number = {12-005},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
20.
Millman, David L
Degree-Driven Design of Geometric Algorithms for Point Location, Proximity, and Volume Calculation Technical Report
no. 12-004, 2012.
@techreport{CS12-004,
title = {Degree-Driven Design of Geometric Algorithms for Point Location, Proximity, and Volume Calculation},
author = {David L Millman},
url = {https://techreports.cs.unc.edu/papers/12-004.pdf},
year = {2012},
date = {2012-12-21},
urldate = {1970-01-01},
number = {12-004},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
Technical Reports produced by the Department of Computer Science at the University of North Carolina at Chapel Hill.
NOTE: All reports to which we have access have been made available here. Contact techrep@cs.unc.edu if you have problems downloading one of these reports.