Learning and Exploiting Camera Geometry for Computer Vision

<p>This work explores the use of statistical methods in describing and estimating camera poses, as well as the information feedback loop between camera pose and object detection. Surging development in robotics and computer vision has pushed the need for algorithms that infer, understand, and utilize information about the position and orientation of the sensor platforms when observing and/or interacting with their environment.</p><p>The first contribution of this thesis is the development of a set of statistical tools for representing and estimating the uncertainty in object poses. A distribution for representing the joint uncertainty over multiple object positions and orientations is described, called the mirrored normal-Bingham distribution. This distribution generalizes both the normal distribution in Euclidean space, and the Bingham distribution on the unit hypersphere. It is shown to inherit many of the convenient properties of these special cases: it is the maximum-entropy distribution with fixed second moment, and there is a generalized Laplace approximation whose result is the mirrored normal-Bingham distribution. This distribution and approximation method are demonstrated by deriving the analytical approximation to the wrapped-normal distribution. Further, it is shown how these tools can be used to represent the uncertainty in the result of a bundle adjustment problem.</p><p>Another application of these methods is illustrated as part of a novel camera pose estimation algorithm based on object detections. The autocalibration task is formulated as a bundle adjustment problem using prior distributions over the 3D points to enforce the objects' structure and their relationship with the scene geometry. This framework is very flexible and enables the use of off-the-shelf computational tools to solve specialized autocalibration problems. Its performance is evaluated using a pedestrian detector to provide head and foot location observations, and it proves much faster and potentially more accurate than existing methods.</p><p>Finally, the information feedback loop between object detection and camera pose estimation is closed by utilizing camera pose information to improve object detection in scenarios with significant perspective warping. Methods are presented that allow the inverse perspective mapping traditionally applied to images to be applied instead to features computed from those images. For the special case of HOG-like features, which are used by many modern object detection systems, these methods are shown to provide substantial performance benefits over unadapted detectors while achieving real-time frame rates, orders of magnitude faster than comparable image warping methods.</p><p>The statistical tools and algorithms presented here are especially promising for mobile cameras, providing the ability to autocalibrate and adapt to the camera pose in real time. In addition, these methods have wide-ranging potential applications in diverse areas of computer vision, robotics, and imaging.</p>

Studying Recommender Systems to Enhance Distributed Computing Schedulers

<p>Distributed Computing frameworks belong to a class of programming models that allow developers to</p><p> launch workloads on large clusters of machines. Due to the dramatic increase in the volume of</p><p> data gathered by ubiquitous computing devices, data analytic workloads have become a common</p><p> case among distributed computing applications, making Data Science an entire field of</p><p> Computer Science. We argue that Data Scientist's concern lays in three main components: a dataset,</p><p> a sequence of operations they wish to apply on this dataset, and some constraint they may have</p><p> related to their work (performances, QoS, budget, etc). However, it is actually extremely</p><p> difficult, without domain expertise, to perform data science. One need to select the right amount</p><p> and type of resources, pick up a framework, and configure it. Also, users are often running their</p><p> application in shared environments, ruled by schedulers expecting them to specify precisely their resource</p><p> needs. Inherent to the distributed and concurrent nature of the cited frameworks, monitoring and </p><p> profiling are hard, high dimensional problems that block users from making the right</p><p> configuration choices and determining the right amount of resources they need. Paradoxically, the </p><p> system is gathering a large amount of monitoring data at runtime, which remains unused.</p><p> In the ideal abstraction we envision for data scientists, the system is adaptive, able to exploit</p><p> monitoring data to learn about workloads, and process user requests into a tailored execution</p><p> context. In this work, we study different techniques that have been used to make steps toward</p><p> such system awareness, and explore a new way to do so by implementing machine learning</p><p> techniques to recommend a specific subset of system configurations for Apache Spark applications.</p><p> Furthermore, we present an in depth study of Apache Spark executors configuration, which highlight</p><p> the complexity in choosing the best one for a given workload.</p>

Algorithms for Geometric Matching, Clustering, and Covering

<p>With the popularization of GPS-enabled devices such as mobile phones, location data are becoming available at an unprecedented scale. The locations may be collected from many different sources such as vehicles moving around a city, user check-ins in social networks, and geo-tagged micro-blogging photos or messages. Besides the longitude and latitude, each location record may also have a timestamp and additional information such as the name of the location. Time-ordered sequences of these locations form trajectories, which together contain useful high-level information about people's movement patterns.</p><p>The first part of this thesis focuses on a few geometric problems motivated by the matching and clustering of trajectories. We first give a new algorithm for computing a matching between a pair of curves under existing models such as dynamic time warping (DTW). The algorithm is more efficient than standard dynamic programming algorithms both theoretically and practically. We then propose a new matching model for trajectories that avoids the drawbacks of existing models. For trajectory clustering, we present an algorithm that computes clusters of subtrajectories, which correspond to common movement patterns. We also consider trajectories of check-ins, and propose a statistical generative model, which identifies check-in clusters as well as the transition patterns between the clusters. </p><p>The second part of the thesis considers the problem of covering shortest paths in a road network, motivated by an EV charging station placement problem. More specifically, a subset of vertices in the road network are selected to place charging stations so that every shortest path contains enough charging stations and can be traveled by an EV without draining the battery. We first introduce a general technique for the geometric set cover problem. This technique leads to near-linear-time approximation algorithms, which are the state-of-the-art algorithms for this problem in either running time or approximation ratio. We then use this technique to develop a near-linear-time algorithm for this</p><p>shortest-path cover problem.</p>

Distributed Feature Selection in Large n and Large p Regression Problems

<p>Fitting statistical models is computationally challenging when the sample size or the dimension of the dataset is huge. An attractive approach for down-scaling the problem size is to first partition the dataset into subsets and then fit using distributed algorithms. The dataset can be partitioned either horizontally (in the sample space) or vertically (in the feature space), and the challenge arise in defining an algorithm with low communication, theoretical guarantees and excellent practical performance in general settings. For sample space partitioning, I propose a MEdian Selection Subset AGgregation Estimator ({\em message}) algorithm for solving these issues. The algorithm applies feature selection in parallel for each subset using regularized regression or Bayesian variable selection method, calculates the `median' feature inclusion index, estimates coefficients for the selected features in parallel for each subset, and then averages these estimates. The algorithm is simple, involves very minimal communication, scales efficiently in sample size, and has theoretical guarantees. I provide extensive experiments to show excellent performance in feature selection, estimation, prediction, and computation time relative to usual competitors.</p><p>While sample space partitioning is useful in handling datasets with large sample size, feature space partitioning is more effective when the data dimension is high. Existing methods for partitioning features, however, are either vulnerable to high correlations or inefficient in reducing the model dimension. In the thesis, I propose a new embarrassingly parallel framework named {\em DECO} for distributed variable selection and parameter estimation. In {\em DECO}, variables are first partitioned and allocated to m distributed workers. The decorrelated subset data within each worker are then fitted via any algorithm designed for high-dimensional problems. We show that by incorporating the decorrelation step, DECO can achieve consistent variable selection and parameter estimation on each subset with (almost) no assumptions. In addition, the convergence rate is nearly minimax optimal for both sparse and weakly sparse models and does NOT depend on the partition number m. Extensive numerical experiments are provided to illustrate the performance of the new framework.</p><p>For datasets with both large sample sizes and high dimensionality, I propose a new "divided-and-conquer" framework {\em DEME} (DECO-message) by leveraging both the {\em DECO} and the {\em message} algorithm. The new framework first partitions the dataset in the sample space into row cubes using {\em message} and then partition the feature space of the cubes using {\em DECO}. This procedure is equivalent to partitioning the original data matrix into multiple small blocks, each with a feasible size that can be stored and fitted in a computer in parallel. The results are then synthezied via the {\em DECO} and {\em message} algorithm in a reverse order to produce the final output. The whole framework is extremely scalable.</p>

On the fly type specialization without type analysis

Les langages de programmation typ��s dynamiquement tels que JavaScript et Python repoussent la v��rification de typage jusqu���au moment de l���ex��cution. Afin d���optimiser la performance de ces langages, les impl��mentations de machines virtuelles pour langages dynamiques doivent tenter d�����liminer les tests de typage dynamiques redondants. Cela se fait habituellement en utilisant une analyse d���inf��rence de types. Cependant, les analyses de ce genre sont souvent co��teuses et impliquent des compromis entre le temps de compilation et la pr��cision des r��sultats obtenus. Ceci a conduit �� la conception d���architectures de VM de plus en plus complexes. Nous proposons le versionnement paresseux de blocs de base, une technique de compilation �� la vol��e simple qui ��limine efficacement les tests de typage dynamiques redondants sur les chemins d���ex��cution critiques. Cette nouvelle approche g��n��re paresseusement des versions sp��cialis��es des blocs de base tout en propageant de l���information de typage contextualis��e. Notre technique ne n��cessite pas l���utilisation d���analyses de programme co��teuses, n���est pas contrainte par les limitations de pr��cision des analyses d���inf��rence de types traditionnelles et ��vite la complexit�� des techniques d���optimisation sp��culatives. Trois extensions sont apport��es au versionnement de blocs de base afin de lui donner des capacit��s d���optimisation interproc��durale. Une premi��re extension lui donne la possibilit�� de joindre des informations de typage aux propri��t��s des objets et aux variables globales. Puis, la sp��cialisation de points d���entr��e lui permet de passer de l���information de typage des fonctions appellantes aux fonctions appell��es. Finalement, la sp��cialisation des continuations d���appels permet de transmettre le type des valeurs de retour des fonctions appell��es aux appellants sans co��t dynamique. Nous d��montrons empiriquement que ces extensions permettent au versionnement de blocs de base d�����liminer plus de tests de typage dynamiques que toute analyse d���inf��rence de typage statique.

Application of computer-aided tomography techniques to visualize kelp holdfast structure reveals the importance of habitat complexity for supporting marine biodiversity

Ecosystem engineers that increase habitat complexity are keystone species in marine systems, increasing shelter and niche availability, and therefore biodiversity. For example, kelp holdfasts form intricate structures and host the largest number of organisms in kelp ecosystems. However, methods that quantify 3D habitat complexity have only seldom been used in marine habitats, and never in kelp holdfast communities. This study investigated the role of kelp holdfasts (Laminaria hyperborea) in supporting benthic faunal biodiversity. Computer-aided tomography (CT-) scanning was used to quantify the three-dimensional geometrical complexity of holdfasts, including volume, surface area and surface fractal dimension (FD). Additionally, the number of haptera, number of haptera per unit of volume, and age of kelps were estimated. These measurements were compared to faunal biodiversity and community structure, using partial least-squares regression and multivariate ordination. Holdfast volume explained most of the variance observed in biodiversity indices, however all other complexity measures also strongly contributed to the variance observed. Multivariate ordinations further revealed that surface area and haptera per unit of volume accounted for the patterns observed in faunal community structure. Using 3D image analysis, this study makes a strong contribution to elucidate quantitative mechanisms underlying the observed relationship between biodiversity and habitat complexity. Furthermore, the potential of CT-scanning as an ecological tool is demonstrated, and a methodology for its use in future similar studies is established. Such spatially resolved imager analysis could help identify structurally complex areas as biodiversity hotspots, and may support the prioritization of areas for conservation.

Application of computer-aided tomography techniques to visualize kelp holdfast structure reveals the importance of habitat complexity for supporting marine biodiversity

Ecosystem engineers that increase habitat complexity are keystone species in marine systems, increasing shelter and niche availability, and therefore biodiversity. For example, kelp holdfasts form intricate structures and host the largest number of organisms in kelp ecosystems. However, methods that quantify 3D habitat complexity have only seldom been used in marine habitats, and never in kelp holdfast communities. This study investigated the role of kelp holdfasts (Laminaria hyperborea) in supporting benthic faunal biodiversity. Computer-aided tomography (CT-) scanning was used to quantify the three-dimensional geometrical complexity of holdfasts, including volume, surface area and surface fractal dimension (FD). Additionally, the number of haptera, number of haptera per unit of volume, and age of kelps were estimated. These measurements were compared to faunal biodiversity and community structure, using partial least-squares regression and multivariate ordination. Holdfast volume explained most of the variance observed in biodiversity indices, however all other complexity measures also strongly contributed to the variance observed. Multivariate ordinations further revealed that surface area and haptera per unit of volume accounted for the patterns observed in faunal community structure. Using 3D image analysis, this study makes a strong contribution to elucidate quantitative mechanisms underlying the observed relationship between biodiversity and habitat complexity. Furthermore, the potential of CT-scanning as an ecological tool is demonstrated, and a methodology for its use in future similar studies is established. Such spatially resolved imager analysis could help identify structurally complex areas as biodiversity hotspots, and may support the prioritization of areas for conservation.

Assessment of the EEE-4 oral test: a discourse analysis based on complex networks.

With the development of information technology, the theory and methodology of complex network has been introduced to the language research, which transforms the system of language in a complex networks composed of nodes and edges for the quantitative analysis about the language structure. The development of dependency grammar provides theoretical support for the construction of a treebank corpus, making possible a statistic analysis of complex networks. This paper introduces the theory and methodology of the complex network and builds dependency syntactic networks based on the treebank of speeches from the EEE-4 oral test. According to the analysis of the overall characteristics of the networks, including the number of edges, the number of the nodes, the average degree, the average path length, the network centrality and the degree distribution, it aims to find in the networks potential difference and similarity between various grades of speaking performance. Through clustering analysis, this research intends to prove the network parameters��� discriminating feature and provide potential reference for scoring speaking performance.

A Proposed Framework for Simulation Based Learning of Inheritance

Different types of serious games have been used in elucidating computer science areas such as computer games, mobile games, Lego-based games, virtual worlds and webbased games. Different evaluation techniques have been conducted like questionnaires, interviews, discussions and tests. Simulation have been widely used in computer science as a motivational and interactive learning tool. This paper aims to evaluate the possibility of successful implementation of simulation in computer programming modules. A framework is proposed to measure the impact of serious games on enhancing students understanding of key computer science concepts. Experiments will be held on the EEECS of Queen���s University Belfast students to test the framework and attain results.

Jigsaw Java: Intelligent, Automated Marking

The popularity of Computing degrees in the UK has been increasing significantly over the past number of years. In Northern Ireland, from 2007 to 2015, there has been a 40% increase in acceptances to Computer Science degrees with England seeing a 60% increase over the same period (UCAS, 2016). However, this is tainted as Computer Science degrees also continue to maintain the highest dropout rates. <br/>In Queen���s University Belfast we currently have a Level 1 intake of over 400 students across a number of computing pathways. Our drive as staff is to empower and motivate the students to fully engage with the course content. All students take a Java programming module the aim of which is to provide an understanding of the basic principles of object-oriented design. In order to assess these skills, we have developed Jigsaw Java as an innovative assessment tool offering intelligent, semi-supervised automated marking of code. <br/>Jigsaw Java allows students to answer programming questions using a drag-and-drop interface to place code fragments into position. Their answer is compared to the sample solution and if it matches, marks are allocated accordingly. However, if a match is not found then the corresponding code is executed using sample data to determine if its logic is acceptable. If it is, the solution is flagged to be checked by staff and if satisfactory is saved as an alternative solution. This means that appropriate marks can be allocated and should another student have submitted the same placement of code fragments this does not need to be executed or checked again. Rather the system now knows how to assess it. <br/>Jigsaw Java is also able to consider partial marks dependent on code placement and will ���learn��� over time. Given the number of students, Jigsaw Java will improve the consistency and timeliness of marking.<br/>

The Loop-of-Stencil-Reduce Paradigm

<p>In this paper we advocate the Loop-of-stencil-reduce pattern as a way to simplify the parallel programming of heterogeneous platforms (multicore+GPUs). Loop-of-Stencil-reduce is general enough to subsume map, reduce, map-reduce, stencil, stencil-reduce, and, crucially, their usage in a loop. It transparently targets (by using OpenCL) combinations of CPU cores and GPUs, and it makes it possible to simplify the deployment of a single stencil computation kernel on different GPUs. The paper discusses the implementation of Loop-of-stencil-reduce within the FastFlow parallel framework, considering a simple iterative data-parallel application as running example (Game of Life) and a highly effective parallel filter for visual data restoration to assess performance. Thanks to the high-level design of the Loop-of-stencil-reduce, it was possible to run the filter seamlessly on a multicore machine, on multi-GPUs, and on both.</p>

Detecting stealthy attacks: Efficient monitoring of suspicious activities on computer networks

<p>Stealthy attackers move patiently through computer networks - taking days, weeks or months to accomplish their objectives in order to avoid detection. As networks scale up in size and speed, monitoring for such attack attempts is increasingly a challenge. This paper presents an efficient monitoring technique for stealthy attacks. It investigates the feasibility of proposed method under number of different test cases and examines how design of the network affects the detection. A methodological way for tracing anonymous stealthy activities to their approximate sources is also presented. The Bayesian fusion along with traffic sampling is employed as a data reduction method. The proposed method has the ability to monitor stealthy activities using 10-20% size sampling rates without degrading the quality of detection.</p>

Proposing the deep dynamic Bayesian network as a future computer based medical system

<p>The development of new learning models has been of great importance throughout recent years, with a focus on creating advances in the area of deep learning. Deep learning was first noted in 2006, and has since become a major area of research in a number of disciplines. This paper will delve into the area of deep learning to present its current limitations and provide a new idea for a fully integrated deep and dynamic probabilistic system. The new model will be applicable to a vast number of areas initially focusing on applications into medical image analysis with an overall goal of utilising this approach for prediction purposes in computer based medical systems.</p>

Model-driven development of sensor network applications with optimization of non-functional constraints

Wireless sensor networks (WSNs) differ from conventional distributed systems in many aspects. The resource limitation of sensor nodes, the ad-hoc communication and topology of the network, coupled with an unpredictable deployment environment are difficult non-functional constraints that must be carefully taken into account when developing software systems for a WSN. Thus, more research needs to be done on designing, implementing and maintaining software for WSNs. This thesis aims to contribute to research being done in this area by presenting an approach to WSN application development that will improve the reusability, flexibility, and maintainability of the software. Firstly, we present a programming model and software architecture aimed at describing WSN applications, independently of the underlying operating system and hardware. The proposed architecture is described and realized using the Model-Driven Architecture (MDA) standard in order to achieve satisfactory levels of encapsulation and abstraction when programming sensor nodes. Besides, we study different non-functional constrains of WSN application and propose two approaches to optimize the application to satisfy these constrains. A real prototype framework was built to demonstrate the developed solutions in the thesis. The framework implemented the programming model and the multi-layered software architecture as components. A graphical interface, code generation components and supporting tools were also included to help developers design, implement, optimize, and test the WSN software. Finally, we evaluate and critically assess the proposed concepts. Two case studies are provided to support the evaluation. The first case study, a framework evaluation, is designed to assess the ease at which novice and intermediate users can develop correct and power efficient WSN applications, the portability level achieved by developing applications at a high-level of abstraction, and the estimated overhead due to usage of the framework in terms of the footprint and executable code size of the application. In the second case study, we discuss the design, implementation and optimization of a real-world application named TempSense, where a sensor network is used to monitor the temperature within an area.

Maude Object-Oriented Action Tool

MAIDL, Andr�� Murbach; CARVILHE, Claudio; MUSICANTE, Martin A. Maude Object-Oriented Action Tool. Electronic Notes in Theoretical Computer Science. [S.l:s.n], 2008.