Valutazione sperimentale di metodologie di rettificazione e impatto su algoritmi di visione stereo

Ricavare informazioni dalla realtà circostante è un obiettivo molto importante dell'informatica moderna, in modo da poter progettare robot, veicoli a guida autonoma, sistemi di riconoscimento e tanto altro. La computer vision è la parte dell'informatica che se ne occupa e sta sempre più prendendo piede. Per raggiungere tale obiettivo si utilizza una pipeline di visione stereo i cui passi di rettificazione e generazione di mappa di disparità sono oggetto di questa tesi. In particolare visto che questi passi sono spesso affidati a dispositivi hardware dedicati (come le FPGA) allora si ha la necessità di utilizzare algoritmi che siano portabili su questo tipo di tecnologia, dove le risorse sono molto minori. Questa tesi mostra come sia possibile utilizzare tecniche di approssimazione di questi algoritmi in modo da risparmiare risorse ma che che garantiscano comunque ottimi risultati.

How do Small Players Deduce Beliefs about Uncertainty? A Look at Texas Shale Oil Investments

Thesis (Ph.D.)--University of Washington, 2016-08

Continuous Delivery : Utmaningar ett förvaltningsuppdrag kan stå inför vid en övergång till Continuous Delivery

Continuous delivery (CD) is a software engineering approach where the focus lays on creating a short delivery cycle by automating parts of the deployment pipeline which includes build, deploy-, test and release process. CD is based on that during development should be possible to always automatically generate a release based on the source code in its current state. One of CD's many advantages is that through continuous releases it allows you to get a quick feedback loop leading to faster and more efficient implementation of new functions, at the same time fixing errors. Although CD has many advantages, there are also several challenges a maintenance management project must manage in the transition to CD. These challenges may differ depending on the maturity level for a maintenance management project and what strengths and weaknesses the project has. Our research question was: "What challenges can a maintenance management project face in transition to Continuous delivery?" The purpose of this study is to describe Continuous delivery and the challenges a maintenance management project may face during a transition to Continuous delivery. A descriptive case study has been carried out with the data collection methods of interviews and documents. A situation analysis was created based on the collected data in a shape of a process model that represent the maintenance management projects release process. The processmodel was used as the basis of SWOT analysis and analysis by Rehn et al's Maturity Model. From these analyzes we found challenges of a maintenance management project may face in the transition to CD. The challenges are about customers and the management's attitude towards a transition to CD. But the biggest challenge is about automation of the deployment pipeline steps.

Feature-based camera pose estimation

The estimating of the relative orientation and position of a camera is one of the integral topics in the field of computer vision. The accuracy of a certain Finnish technology company’s traffic sign inventory and localization process can be improved by utilizing the aforementioned concept. The company’s localization process uses video data produced by a vehicle installed camera. The accuracy of estimated traffic sign locations depends on the relative orientation between the camera and the vehicle. This thesis proposes a computer vision based software solution which can estimate a camera’s orientation relative to the movement direction of the vehicle by utilizing video data. The task was solved by using feature-based methods and open source software. When using simulated data sets, the camera orientation estimates had an absolute error of 0.31 degrees on average. The software solution can be integrated to be a part of the traffic sign localization pipeline of the company in question.

School-Based Referrals to the Juvenile Courts: Prevalence and Characteristics in Several States

Effective school discipline practices are essential to keeping schools safe and creating an optimal learning environment. However, the overreliance of exclusionary discipline often removes students from the school setting and deprives them of the opportunity to learn. Previous research has suggested that students are being introduced to the juvenile justice system through the use of school-based juvenile court referrals. In 2011, approximately 1.2 million delinquency cases were referred to the juvenile courts in the United States. Preliminary evidence suggests that an increasing number of these referrals have originated in the schools. This study investigated school-based referrals to the juvenile courts as an element of the School-to-Prison Pipeline (StPP). The likelihood of school-based juvenile court referrals and rate of dismissal of these referrals was examined in several states using data from the National Juvenile Court Data Archives. In addition, the study examined race and special education status as predictors of school-based juvenile court referrals. Descriptive statistics, logistic regression and odds ratio, were used to analyze the data, make conclusions based on the findings and recommend appropriate school discipline practices.

A Study of the State of the Art of Process Planning for Additive Manufacturing

In the manufacturing industry the term Process Planning (PP) is concerned with determining the sequence of individual manufacturing operations needed to produce a given part or product with a certain machine. In this technical report we propose a preliminary analysis of scientific literature on the topic of process planning for Additive Manufacturing (AM) technologies (i.e. 3D printing). We observe that the process planning for additive manufacturing processes consists of a small set of standard operations (repairing, orientation, supports, slicing and toolpath generation). We analyze each of them in order to emphasize the most critical aspects of the current pipeline as well as highlight the future challenges for this emerging manufacturing technology.

Discovery driven analysis on semi-structured text data

Discovery Driven Analysis (DDA) is a common feature of OLAP technology to analyze structured data. In essence, DDA helps analysts to discover anomalous data by highlighting 'unexpected' values in the OLAP cube. By giving indications to the analyst on what dimensions to explore, DDA speeds up the process of discovering anomalies and their causes. However, Discovery Driven Analysis (and OLAP in general) is only applicable on structured data, such as records in databases. We propose a system to extend DDA technology to semi-structured text documents, that is, text documents with a few structured data. Our system pipeline consists of two stages: first, the text part of each document is structured around user specified dimensions, using semi-PLSA algorithm; then, we adapt DDA to these fully structured documents, thus enabling DDA on text documents. We present some applications of this system in OLAP analysis and show how scalability issues are solved. Results show that our system can handle reasonable datasets of documents, in real time, without any need for pre-computation.

New developments in UAV remote sensing for pest management and the implications for cotton

Efficient crop monitoring and pest damage assessments are key to protecting the Australian agricultural industry and ensuring its leading position internationally. An important element in pest detection is gathering reliable crop data frequently and integrating analysis tools for decision making. Unmanned aerial systems are emerging as a cost-effective solution to a number of precision agriculture challenges. An important advantage of this technology is it provides a non-invasive aerial sensor platform to accurately monitor broad acre crops. In this presentation, we will give an overview on how unmanned aerial systems and machine learning can be combined to address crop protection challenges. A recent 2015 study on insect damage in sorghum will illustrate the effectiveness of this methodology. A UAV platform equipped with a high-resolution camera was deployed to autonomously perform a flight pattern over the target area. We describe the image processing pipeline implemented to create a georeferenced orthoimage and visualize the spatial distribution of the damage. An image analysis tool has been developed to minimize human input requirements. The computer program is based on a machine learning algorithm that automatically creates a meaningful partition of the image into clusters. Results show the algorithm delivers decision boundaries that accurately classify the field into crop health levels. The methodology presented in this paper represents a venue for further research towards automated crop protection assessments in the cotton industry, with applications in detecting, quantifying and monitoring the presence of mealybugs, mites and aphid pests.

Quasispecies dynamics and treatment outcome during early hepatitis C infection in a cohort of HIV-infected men

Hepatitis C virus (HCV) is emerging as one of the leading causes of morbidity and mortality in individuals infected with HIV and has overtaken AIDS-defining illnesses as a cause of death in HIV patient populations who have access to highly active antiretroviral therapy. For many years, the clonal analysis was the reference method for investigating viral diversity. In this thesis, a next generation sequencing (NGS) approach was developed using 454 pyrosequencing and Illumina-based technology. A sequencing pipeline was developed using two different NGS approaches, nested PCR, and metagenomics. The pipeline was used to study the viral populations in the sera of HCV-infected patients from a unique cohort of 160 HIV-positive patients with early HCV infection. These pipelines resulted in an improved understanding of HCV quasispecies dynamics, especially regarding studying response to treatment. Low viral diversity at baseline correlated with sustained virological response (SVR) while high viral diversity at baseline was associated with treatment failure. The emergence of new viral strains following treatment failure was most commonly associated with emerging dominance of pre-existing minority variants rather than re-infection. In the new era of direct-acting antivirals, next generation sequencing technologies are the most promising tool for identifying minority variants present in the HCV quasispecies populations at baseline. In this cohort, several mutations conferring resistance were detected in genotype 1a treatment-naïve patients. Further research into the impact of baseline HCV variants on SVR rates should be carried out in this population. A clearer understanding of the properties of viral quasispecies would enable clinicians to make improved treatment choices for their patients.

Virtual-Build-to-Order as a Mass Customization Order Fulfilment Model

Virtual-build-to-order (VBTO) is a form of order fulfilment system in which the producer has the ability to search across the entire pipeline of finished stock, products in production and those in the production plan, in order to find the best product for a customer. It is a system design that is attractive to Mass Customizers, such as those in the automotive sector, whose manufacturing lead time exceeds their customers' tolerable waiting times, and for whom the holding of partly-finished stocks at a fixed decoupling point is unattractive or unworkable. This paper describes and develops the operational concepts that underpin VBTO, in particular the concepts of reconfiguration flexibility and customer aversion to waiting. Reconfiguration is the process of changing a product's specification at any point along the order fulfilment pipeline. The extent to which an order fulfilment system is flexible or inflexible reveals itself in the reconfiguration cost curve, of which there are four basic types. The operational features of the generic VBTO system are described and simulation is used to study its behaviour and performance. The concepts of reconfiguration flexibility and floating decoupling point are introduced and discussed.

Fundamental behaviour of Virtual-Build-to-Order systems

Virtual-Build-to-Order (VBTO) is an emerging order fulfilment system within the automotive sector that is intended to improve fulfilment performance by taking advantage of integrated information systems. The primary innovation in VBTO systems is the ability to make available all unsold products that are in the production pipeline to all customers. In a conventional system the pipeline is inaccessible and a customer can be fulfilled by a product from stock or having a product Built-to-Order (BTO), whereas in a VBTO system a customer can be fulfilled by a product from stock, by being allocated a product in the pipeline, or by a build-to-order product. Simulation is used to investigate and profile the fundamental behaviour of the basic VBTO system and to compare it to a Conventional system. A predictive relationship is identified, between the proportions of customers fulfilled through each mechanism and the ratio of product variety / pipeline length. The simulations reveal that a VBTO system exhibits inherent behaviour that alters the stock mix and levels, leading to stock levels being higher than in an equivalent conventional system at certain variety / pipeline ratios. The results have implications for the design and management of order fulfilment systems in sectors such as automotive where VBTO is a viable operational model.

MediPlEx - a tool to combine in silico & experimental gene expression profiles of the model legume Medicago truncatula

Background: Expressed Sequence Tags (ESTs) are in general used to gain a first insight into gene activities from a species of interest. Subsequently, and typically based on a combination of EST and genome sequences, microarray-based expression analyses are performed for a variety of conditions. In some cases, a multitude of EST and microarray experiments are conducted for one species, covering different tissues, cell states, and cell types. Under these circumstances, the challenge arises to combine results derived from the different expression profiling strategies, with the goal to uncover novel information on the basis of the integrated datasets. Findings: Using our new analysis tool, MediPlEx (MEDIcago truncatula multiPLe EXpression analysis), expression data from EST experiments, oligonucleotide microarrays and Affymetrix GeneChips® can be combined and analyzed, leading to a novel approach to integrated transcriptome analysis. We have validated our tool via the identification of a set of well-characterized AM-specific and AM-induced marker genes, identified by MediPlEx on the basis of in silico and experimental gene expression profiles from roots colonized with AM fungi. Conclusions: MediPlEx offers an integrated analysis pipeline for different sets of expression data generated for the model legume Medicago truncatula. As expected, in silico and experimental gene expression data that cover the same biological condition correlate well. The collection of differentially expressed genes identified via MediPlEx provides a starting point for functional studies in plant mutants.

Integrated visual perception architecture for robotic clothes perception and manipulation

This thesis proposes a generic visual perception architecture for robotic clothes perception and manipulation. This proposed architecture is fully integrated with a stereo vision system and a dual-arm robot and is able to perform a number of autonomous laundering tasks. Clothes perception and manipulation is a novel research topic in robotics and has experienced rapid development in recent years. Compared to the task of perceiving and manipulating rigid objects, clothes perception and manipulation poses a greater challenge. This can be attributed to two reasons: firstly, deformable clothing requires precise (high-acuity) visual perception and dexterous manipulation; secondly, as clothing approximates a non-rigid 2-manifold in 3-space, that can adopt a quasi-infinite configuration space, the potential variability in the appearance of clothing items makes them difficult to understand, identify uniquely, and interact with by machine. From an applications perspective, and as part of EU CloPeMa project, the integrated visual perception architecture refines a pre-existing clothing manipulation pipeline by completing pre-wash clothes (category) sorting (using single-shot or interactive perception for garment categorisation and manipulation) and post-wash dual-arm flattening. To the best of the author’s knowledge, as investigated in this thesis, the autonomous clothing perception and manipulation solutions presented here were first proposed and reported by the author. All of the reported robot demonstrations in this work follow a perception-manipulation method- ology where visual and tactile feedback (in the form of surface wrinkledness captured by the high accuracy depth sensor i.e. CloPeMa stereo head or the predictive confidence modelled by Gaussian Processing) serve as the halting criteria in the flattening and sorting tasks, respectively. From scientific perspective, the proposed visual perception architecture addresses the above challenges by parsing and grouping 3D clothing configurations hierarchically from low-level curvatures, through mid-level surface shape representations (providing topological descriptions and 3D texture representations), to high-level semantic structures and statistical descriptions. A range of visual features such as Shape Index, Surface Topologies Analysis and Local Binary Patterns have been adapted within this work to parse clothing surfaces and textures and several novel features have been devised, including B-Spline Patches with Locality-Constrained Linear coding, and Topology Spatial Distance to describe and quantify generic landmarks (wrinkles and folds). The essence of this proposed architecture comprises 3D generic surface parsing and interpretation, which is critical to underpinning a number of laundering tasks and has the potential to be extended to other rigid and non-rigid object perception and manipulation tasks. The experimental results presented in this thesis demonstrate that: firstly, the proposed grasp- ing approach achieves on-average 84.7% accuracy; secondly, the proposed flattening approach is able to flatten towels, t-shirts and pants (shorts) within 9 iterations on-average; thirdly, the proposed clothes recognition pipeline can recognise clothes categories from highly wrinkled configurations and advances the state-of-the-art by 36% in terms of classification accuracy, achieving an 83.2% true-positive classification rate when discriminating between five categories of clothes; finally the Gaussian Process based interactive perception approach exhibits a substantial improvement over single-shot perception. Accordingly, this thesis has advanced the state-of-the-art of robot clothes perception and manipulation.

Filling in gaps of Drosophila melanogaster urate degradation metabolic pathway using metabolomics approaches: towards the core metabolome of the fruit fly

The primary goal of systems biology is to integrate complex omics data, and data obtained from traditional experimental studies in order to provide a holistic understanding of organismal function. One way of achieving this aim is to generate genome-scale metabolic models (GEMs), which contain information on all metabolites, enzyme-coding genes, and biochemical reactions in a biological system. Drosophila melanogaster GEM has not been reconstructed to date. Constraint-free genome-wide metabolic model of the fruit fly has been reconstructed in our lab, identifying gaps, where no enzyme was identified and metabolites were either only produced or consume. The main focus of the work presented in this thesis was to develop a pipeline for efficient gap filling using metabolomics approaches combined with standard reverse genetics methods, using 5-hydroxyisourate hydrolase (5-HIUH) as an example. 5-HIUH plays a role in urate degradation pathway. Inability to degrade urate can lead to inborn errors of metabolism (IEMs) in humans, including hyperuricemia. Based on sequence analysis Drosophila CG30016 gene was hypothesised to encode 5- HIUH. CG30016 knockout flies were examined to identify Malpighian tubules phenotype, and shortened lifespan might reflect kidney disorders in hyperuricemia in humans. Moreover, LC-MS analysis of mutant tubules revealed that CG30016 is involved in purine metabolism, and specifically urate degradation pathway. However, the exact role of the gene has not been identified, and the complete method for gap filling has not been developed. Nevertheless, thanks to the work presented here, we are a step closer towards the development of a gap-filling pipeline in Drosophila melanogaster GEM. Importantly, the areas that require further optimisation were identified and are the focus of future research. Moreover, LC-MS analysis confirmed that tubules rather than the whole fly were more suitable for metabolomics analysis of purine metabolism. Previously, Dow/Davies lab has generated the most complete tissue-specific transcriptomic atlas for Drosophila – FlyAtlas.org, which provides data on gene expression across multiple tissues of adult fly and larva. FlyAtlas revealed that transcripts of many genes are enriched in specific Drosophila tissues, and that it is possible to deduce the functions of individual tissues within the fly. Based on FlyAtlas data, it has become clear that the fly (like other metazoan species) must be considered as a set of tissues, each 2 with its own distinct transcriptional and functional profile. Moreover, it revealed that for about 30% of the genome, reverse genetic methods (i.e. mutation in an unknown gene followed by observation of phenotype) are only useful if specific tissues are investigated. Based on the FlyAtlas findings, we aimed to build a primary tissue-specific metabolome of the fruit fly, in order to establish whether different Drosophila tissues have different metabolomes and if they correspond to tissue-specific transcriptome of the fruit fly (FlyAtlas.org). Different fly tissues have been dissected and their metabolome elucidated using LC-MS. The results confirmed that tissue metabolomes differ significantly from each other and from the whole fly, and that some of these differences can be correlated to the tissue function. The results illustrate the need to study individual tissues as well as the whole organism. It is clear that some metabolites that play an important role in a given tissue might not be detected in the whole fly sample because their abundance is much lower in comparison to other metabolites present in all tissues, which prevent the detection of the tissue-specific compound.

Using interaction data for improving the offline and online evaluation of search engines

This thesis investigates how web search evaluation can be improved using historical interaction data. Modern search engines combine offline and online evaluation approaches in a sequence of steps that a tested change needs to pass through to be accepted as an improvement and subsequently deployed. We refer to such a sequence of steps as an evaluation pipeline. In this thesis, we consider the evaluation pipeline to contain three sequential steps: an offline evaluation step, an online evaluation scheduling step, and an online evaluation step. In this thesis we show that historical user interaction data can aid in improving the accuracy or efficiency of each of the steps of the web search evaluation pipeline. As a result of these improvements, the overall efficiency of the entire evaluation pipeline is increased. Firstly, we investigate how user interaction data can be used to build accurate offline evaluation methods for query auto-completion mechanisms. We propose a family of offline evaluation metrics for query auto-completion that represents the effort the user has to spend in order to submit their query. The parameters of our proposed metrics are trained against a set of user interactions recorded in the search engine’s query logs. From our experimental study, we observe that our proposed metrics are significantly more correlated with an online user satisfaction indicator than the metrics proposed in the existing literature. Hence, fewer changes will pass the offline evaluation step to be rejected after the online evaluation step. As a result, this would allow us to achieve a higher efficiency of the entire evaluation pipeline. Secondly, we state the problem of the optimised scheduling of online experiments. We tackle this problem by considering a greedy scheduler that prioritises the evaluation queue according to the predicted likelihood of success of a particular experiment. This predictor is trained on a set of online experiments, and uses a diverse set of features to represent an online experiment. Our study demonstrates that a higher number of successful experiments per unit of time can be achieved by deploying such a scheduler on the second step of the evaluation pipeline. Consequently, we argue that the efficiency of the evaluation pipeline can be increased. Next, to improve the efficiency of the online evaluation step, we propose the Generalised Team Draft interleaving framework. Generalised Team Draft considers both the interleaving policy (how often a particular combination of results is shown) and click scoring (how important each click is) as parameters in a data-driven optimisation of the interleaving sensitivity. Further, Generalised Team Draft is applicable beyond domains with a list-based representation of results, i.e. in domains with a grid-based representation, such as image search. Our study using datasets of interleaving experiments performed both in document and image search domains demonstrates that Generalised Team Draft achieves the highest sensitivity. A higher sensitivity indicates that the interleaving experiments can be deployed for a shorter period of time or use a smaller sample of users. Importantly, Generalised Team Draft optimises the interleaving parameters w.r.t. historical interaction data recorded in the interleaving experiments. Finally, we propose to apply the sequential testing methods to reduce the mean deployment time for the interleaving experiments. We adapt two sequential tests for the interleaving experimentation. We demonstrate that one can achieve a significant decrease in experiment duration by using such sequential testing methods. The highest efficiency is achieved by the sequential tests that adjust their stopping thresholds using historical interaction data recorded in diagnostic experiments. Our further experimental study demonstrates that cumulative gains in the online experimentation efficiency can be achieved by combining the interleaving sensitivity optimisation approaches, including Generalised Team Draft, and the sequential testing approaches. Overall, the central contributions of this thesis are the proposed approaches to improve the accuracy or efficiency of the steps of the evaluation pipeline: the offline evaluation frameworks for the query auto-completion, an approach for the optimised scheduling of online experiments, a general framework for the efficient online interleaving evaluation, and a sequential testing approach for the online search evaluation. The experiments in this thesis are based on massive real-life datasets obtained from Yandex, a leading commercial search engine. These experiments demonstrate the potential of the proposed approaches to improve the efficiency of the evaluation pipeline.