Browsing by Title

X-ray computed tomography (CT) is widely used in medical imaging and materials science. In this imaging modality, cross-sectional images of a physical object are formed by taking numerous X-ray projections from different angles and then applying a reconstruction algorithm to the measured data. The cross-sectional slices can be used to form a three-dimensional model of the interior structure of the object. CT is a prime example of an inverse problem, in which the aim is to recover an unknown cause from a known effect. CT technology continues to develop, motivated by the desire for increased image quality and spatial resolution in reconstructions. In medical CT, reducing patient dose is a major goal. The branch of CT known as X-ray microtomography (micro-CT) produces reconstructions with spatial resolutions in the micrometer range. Micro-CT has been practiced at the University of Helsinki since 2008. The research projects are often interdisciplinary, combining physics with fields such as biosciences, paleontology, geology, geophysics, metallurgy and food technology. This thesis documents the design and construction of a new X-ray imaging system for computed tomography. The system is a cone beam micro-CT scanner intended for teaching and research in inverse problems and X-ray physics. The scanner consists of a molybdenum target X-ray tube, a sample manipulator, and a flat panel detector, and it is built inside a radiation shielding cabinet. Measurements were made for calibrating the measurement geometry and for testing reconstruction quality. Two-dimensional reconstructions of various samples were computed using the plane which passes through the X-ray point source and is perpendicular to the axis of rotation. This central plane of the cone beam reduces to fan beam geometry. All reconstructions were computed using the filtered backprojection (FBP) algorithm, which is the industry standard. Tomographic reconstructions of high quality were obtained from the measurements. The results show that the imaging system is well suited for CT and the study of reconstruction algorithms.

The role of geographic thinking is essential in tackling topical challenges such as climate crisis, biodiversity loss and sustainable production of food. One powerful tool that helps to model and analyze these complex geographic phenomena is geographic information systems (GIS). Using GIS as part of geography high school education has many benefits when it is applied intentionally. However, many teachers still struggle to implement GIS in long-term classroom use even if they have gotten previous GIS training and have access to internet-based GIS, easy-access data and easier to use software. There is still a need for further research on how teachers can be supported in GIS education on a practical level. This thesis research aims to find solutions to this need. The research is conducted as design-based research that consists of problem analyses and a cyclic development process where a design solution, a GIS learning activity, is created. Problem analyses showed that combining new and existing knowledge, using multimodal learning environments, and supporting motivation and development of metacognitive skills are important to take into consideration in designing the learning activity. They also examined features that lead to successful GIS teacher training. Conducted interviews revealed that the biggest challenges with GIS education relate to scarcity of time, insufficient technical skills, and training that does not provide practical value. Teachers wished for very practical level support that is efficient timewise and offers them learning materials that are ready for easy classroom use. Based on these findings, a GIS learning activity was designed to answer the common challenges. The practical was tested consecutively by two geography teachers from a collaborative high school. Feedback revealed that the first teacher faced some challenges relating to time management during the lesson but found the activity useful. The second teacher tested the activity after some modifications had been made and the testing was overall successful. Both teachers expressed interest in using the material and the GIS software again in the future. The findings suggest that providing teachers this research-based GIS learning material has potential to support them in GIS education and to remove many common challenges. Some advantages of the practical were offering teachers a web-based GIS with simple user interface, preprocessed data already included in the service and a ready practical that can be completed in one lesson. The theme also supported the national core curriculum which is very valuable in creating new GIS materials for educational use. This study showed that relevant and inquiry-based GIS activities are still needed in high school geography education. It also serves as the first opening for new LUMA Taita -project that promotes international science education collaboration and brings research into schools in an inspiring way.

This master's thesis describes the development and validation of a uniform control interface for drawing robots with ROS2. The robot control software was tasked with taking SVG images as input and producing them as drawings with three different robots. These robots are the Evil Mad Scientist AxiDraw V3/A3, UFACTORY xArm Lite6, and virtual xArm Lite6. The intended use case for the robots and companion control software is experiments studying human perception of the creativity of the drawing robots. The control software was implemented over the course of a little over six months and used a combination of C++ and Python. The design of the software utilizes ROS2 abstractions such as nodes and topics to combine different components of the software. The control software is validated against the given requirements and found to fulfil the main objectives of the project. The most important of these are that the robots successfully draw SVG images, that they do so in a similar time frame, and that these images look very similar. Drawing similarity was tested by scanning images, aligning them using using minimal error, and then comparing them visually after overlaying the images. Comparing aligned images was useful in detecting subtle differences in the drawing similarity of the robots and was used to discover issues with the robot control software. MSE and SSIM were also calculated for a set of these aligned images, allowing for the effect of future changes made to the robot control software to be quantitatively evaluated. Drawing time for the robots was evaluated by measuring the time taken for drawing a set of images. This testing showed that the Axidraw's velocity and acceleration needed to be reduced by 56% so that the xArm Lite6 could draw in similar time.

Deploying machine learning models is found to be a massive issue in the field. DevOps and Continuous Integration and Continuous Delivery (CI/CD) has proven to streamline and accelerate deployments in the field of software development. Creating CI/CD pipelines in software that includes elements of Machine Learning (MLOps) has unique problems, and trail-blazers in the field solve them with the use of proprietary tooling, often offered by cloud providers. In this thesis, we describe the elements of MLOps. We study what the requirements to automate the CI/CD of Machine Learning systems in the MLOps methodology. We study if it is feasible to create a state-of-the-art MLOps pipeline with existing open-source and cloud-native tooling in a cloud provider agnostic way. We designed an extendable and cloud-native pipeline covering most of the CI/CD needs of Machine Learning system. We motivated why Machine Learning systems should be included in the DevOps methodology. We studied what unique challenges machine learning brings to CI/CD pipelines, production environments and monitoring. We analyzed the pipeline’s design, architecture, and implementation details and its applicability and value to Machine Learning projects. We evaluate our solution as a promising MLOps pipeline, that manages to solve many issues of automating a reproducible Machine Learning project and its delivery to production. We designed it as a fully open-source solution that is relatively cloud provider agnostic. Configuring the pipeline to fit the client needs uses easy-to-use declarative configuration languages (YAML, JSON) that require minimal learning overhead.

Self-aware computing is an emerging research area, which aims to solve issues stemming from a combination of increasingly more complex systems and diverse operating conditions. It adapts the key concepts of human self-awareness to the computational context. These concepts are well established in cognitive science, psychology, social psychology and philosophy, but novel in the software engineering context. Self-aware systems are able to acquire information about their environment and their internal state. The obtained information is used to build knowledge through reasoning. This increasing knowledge is used for building an internal learning model, which enables the system to adapt its actions. Self-aware systems can autonomously navigate runtime changes in their goals and environmental conditions, thus enabling a high degree of adaptivity to changing conditions that are difficult to predict at design time. In contrast, traditional systems have to operate with a set of predefined rules that are reliant on the design time knowledge of the designer. This study aims to identify proposed software architecture solutions that are of value for both practitioners and researchers. The study was conducted as a systematic literature review, for which we have developed a repeatable review protocol in order to cover all relevant literature in the area. The review protocol was explicitly defined and applied rigorously. In our review we managed to extract several proposed architecture designs from the reviewed 9 primary studies. These solutions propose several solutions, such as reference architectures, architecture frameworks, and architectural patterns for designing self-aware systems. This study can be used to get an overview of state of the art software architecture designs for self-aware systems. Additionally, this study can provide support for finding future research direction regarding self-aware systems.

Current smartphones have a storage capacity of several gigabytes. More and more information is stored on mobile devices. To meet the challenge of information organization, we turn to desktop search. Users often possess multiple devices, and synchronize (subsets of) information between them. This makes file synchronization more important. This thesis presents Dessy, a desktop search and synchronization framework for mobile devices. Dessy uses desktop search techniques, such as indexing, query and index term stemming, and search relevance ranking. Dessy finds files by their content, metadata, and context information. For example, PDF files may be found by their author, subject, title, or text. EXIF data of JPEG files may be used in finding them. User–defined tags can be added to files to organize and retrieve them later. Retrieved files are ranked according to their relevance to the search query. The Dessy prototype uses the BM25 ranking function, used widely in information retrieval. Dessy provides an interface for locating files for both users and applications. Dessy is closely integrated with the Syxaw file synchronizer, which provides efficient file and metadata synchronization, optimizing network usage. Dessy supports synchronization of search results, individual files, and directory trees. It allows finding and synchronizing files that reside on remote computers, or the Internet. Dessy is designed to solve the problem of efficient mobile desktop search and synchronization, also supporting remote and Internet search. Remote searches may be carried out offline using a downloaded index, or while connected to the remote machine on a weak network. To secure user data, transmissions between the Dessy client and server are encrypted using symmetric encryption. Symmetric encryption keys are exchanged with RSA key exchange. Dessy emphasizes extensibility. Also the cryptography can be extended. Users may tag their files with context tags and control custom file metadata. Adding new indexed file types, metadata fields, ranking methods, and index types is easy. Finding files is done with virtual directories, which are views into the user's files, browseable by regular file managers. On mobile devices, the Dessy GUI provides easy access to the search and synchronization system. This thesis includes results of Dessy synchronization and search experiments, including power usage measurements. Finally, Dessy has been designed with mobility and device constraints in mind. It requires only MIDP 2.0 Mobile Java with FileConnection support, and Java 1.5 on desktop machines.

Bat monitoring is commonly based on audio analysis. By collecting audio recordings from large areas and analysing their content, it is possible estimate distributions of bat species and changes in them. It is easy to collect a large amount of audio recordings by leaving automatic recording units in nature and collecting them later. However, it takes a lot of time and effort to analyse these recordings. Because of that, there is a great need for automatic tools. We developed a program for detecting bat calls automatically from audio recordings. The program is designed for recordings that are collected from Finland with the AudioMoth recording device. Our method is based on a median clipping method that has previously shown promising results in the field of bird song detection. We add several modifications to the basic method in order to make it work well for our purpose. We use real-world field recordings that we have annotated to evaluate the performance of the detector and compare it to two other freely available programs (Kaleidoscope and Bat Detective). Our method showed good results and got the best F2-score in the comparison.

This thesis presents a scalable method for identifying anomalous periods of non-activity in short periodic event sequences. The method is tested with real world point-of-sale (POS) data from grocery retail setting. However, the method can be applied also to other problem domains which produce similar sequential data. The proposed method models the underlying event sequence as a non-homogeneous Poisson process with a piecewise constant rate function. The rate function for the piecewise homogeneous Poisson process can be estimated with a change point detection algorithm that minimises a cost function consisting of the negative Poisson log-likelihood and a penalty term that is linear to the number of change points. The resulting model can be queried for anomalously long periods of time with no events, i.e., waiting times, by defining a threshold below which the waiting time observations are deemed anomalies. The first experimental part of the thesis focuses on model selection, i.e., in finding a penalty value that results in the change point detection algorithm detecting the true changes in the intensity of the arrivals of the events while not reacting to random fluctuations in the data. In the second experimental part the performance of the anomaly detection methodology is measured against stock-out data, which gives an approximate ground truth for the termination of a POS event sequence. The performance of the anomaly detector is found to be subpar in terms of precision and recall, i.e., the true positive rate and the positive predictive value. The number of false positives remains high even with small threshold values. This needs to be taken into account when considering applying the anomaly detection procedure in practice. Nevertheless, the methodology may have practical value in the retail setting, e.g., in guiding the store personnel where to focus their resources in ensuring the availability of the products.

Popularity of mobile instant messaging applications has flourished during the last ten years, and people are using them to exchange private and personal information on daily basis. These applications can be freely installed from online marketplaces, and average users may have several of them installed on their devices. The amount of information available from these messaging applications for a third-party eavesdropper via network traffic analysis has therefore grown significantly as well. Security features of these applications have also been developing over the years, and the communication between the applications and the background server infrastructure nowadays practically always employs encryption. Recently, more advanced end-to-end encryption methods have been developed to hide the content of the exchanged data even from the messaging service providers. Machine learning techniques have successfully been utilized in analyzing encrypted network traffic, and previous research has shown that this approach can effectively be used to detect mobile applications and the actions users are performing in those applications regardless of encryption. While the actual content of the messages and other transferred data cannot be accessed by the eavesdropper, these methods can still lead to serious privacy compromises. This thesis discusses the present state of machine learning-based identification of applications and user actions, how feasible it would be to actually perform such detection in a Wi-Fi network and what kind of privacy concerns would arise.

After the Human Genome Project completed the mapping of human DNA sequence in 2001, a new era began in biological and medical research. The genetic basis of various diseases, such as cancer, could be studied with higher precision than ever before. The map of human genome enabled next-generation sequencing (NGS) techniques and not only DNA sequencing got faster and cheaper to perform, also the amount of data started to increase exponentially. The field of bioinformatics, which combines both computer and life sciences, got a great challenge to handle all the data available and to dig out relevant information out of it. Various tools with heavily enhanced or completely new kinds of algorithms were developed for the demanding task of the analysis of NGS data, which are in the focus of this thesis. For the search of cancer causing mutations, NGS methods enable genome scale studies with the precision of a single molecule. However, the spectacular scale and preciseness of the data offer another challenge – how to distinguish trivial data from the non-trivial, and furthermore, how to separate reliable data from erroneous. The raw data must be put through a pipeline of various processing tools, which organize and humanize the data with the help of the map of human genome. After data processing, the data is feasible for the actual cancer specific analysis, where causative mutations can be hunted down. For this purpose, I have developed an analysis and visualization software, Rikurator, which provides various features and tools to handle the NGS data. Rikurator is designed for comparative analysis of dozens of cancer samples, quality filtering, controlling and visualization to name a few. In addition to tools in data processing pipeline, this thesis will describe features and implementation of Rikurator.