The objective of the project was to examine the possibility to use machine lear- ning without prior knowledge of machine learning or of mathematics and if that is possible also explore the possibility to use machine learning to interpret a picture of a piece of sheet music. The capacity of detecting notes from images of sheet music in the produced model was then compared to an existing pro- gram called Audiveris. The result became a model later used in a comparison with the program Audiveris. The comparison resulted in Audiveris finding al- most 100% of the notes but the new model only being able to detect about a third. The reasons for the big difference were probably: first that Audiveris has been in development for many years and secondly that the training data used for the new model wasn’t enough varied and complicated. To further increase the ability of the new model ́s skill the main point would be the need for more trai- ning data at the training of the model. Even then you need to supply a variation in the data ́s content and its degree of difficulty which is more varied then the first training data. / Målet med projektet har varit att undersöka om det går att applicera maskininlärning utan att ha kunskaper av maskininlärning och matematik sedan tidigare samt om detta är möjligt också undersöka om det går att använda maskininlärning för att tolka en bild av nottecken för musik. Detta användes sedan för att jämföra dess förmåga att upptäcka noter med ett redan existerande program kallat Audiveris. Metoden som användes var att bilder märktes med information om dess innehåll och sedan användes dessa bilder för att träna en maskininlärningsmodell att tolka bilder av nottecken. Sedan gjordes en manuell jämförelse av Audiveris resultat samt resultatet från maskininlärningsmodellen efter att den tränats på de nya notbilderna. Resultatet blev en modell som sedan användes vid jämförelsen med Audiveris. Den jämförelsen resulterade i att Audiveris visade sig bättre än den nytränade modellen då Audiveris hittade nästan 100% av noterna på bilden medan den nya modellen bara hittade cirka 33.3%. Orsaken till den stora skillnaden på upptäckande av nottecken berodde antagligen till stor del på två saker: den första att Audiveris utvecklats under många år och den andra att träningsdatat som användes till den nya modellen inte var tillräckligt varierat och komplicerat. För att vidareutveckla den nya modellens färdighet skulle framför allt mer träningsdata behöva användas vid träningsmomentet. Även då behövde man se till att variationen av material och dess svårighetsgrad blev mer varierat än vid grundmaterialet.

På Karolinska Universitetssjukhuset Huddinge har en stor mängd knäprotesoperationsenkäter staplats på hög under två år. Svaren från dessa ska konverteras till digitalt format så dessa kan lagras i databasen REDCap för att kunna utföra kvalitetskontroll och prospektiv uppföljning i flera år. För att spara arbetstid efterfrågades ett program som kunde läsa in enkätsvaren automatiskt. I detta projekt skapades ett program i MATLAB med målsättning att klara av att läsa enkätmarkeringar samt minst 70 % av personnumren. Dessa personnummer skulle sedan läggas in i ett Excel-ark och övrig svarsdata i ett separat Excel-ark på grund av sekretesslagar. Resultatet blev att programmet inte klarade av att läsa av personnummer och annan handskriven text men klarade att läsa av markerade flervalsfrågor till 90 % säkerhet i just de enkäterna som programmet var designat för. Programmet kan i nuläget användas för smidigare inläsning tillsammans med korrekturläsning. Det rekommenderas dock att fortsätta vidareutveckla programmet innan användning sker. / At Karolinska University Hospital Huddinge, many knee replacement surveys have been piled high for two years. The answers from these must be converted to digital format so that they can be stored in the REDCap database to be able to perform quality control and prospective follow-up for several years. To save working hours, a program that could read the questionnaires automatically was requested. In this project, a program was created in MATLAB with the goal of being able to read questionnaire markings and at least 70% of the social security numbers. These social security numbers were to be written on an Excel sheet and other answer data on a separate Excel sheet due to confidentiality laws. The result was that the program could not handle reading of social security numbers and other handwritten text but managed to read marked multiple-choice questions to 90% certainty in the surveys for which the program was designed. The program can currently be used for easier reading along with proofreading from staff. However, it is recommended to continue to develop the program before using it.

Many industries have been transformed to better perform in today’s digital age. In this project a solution for digitalizing printed sheet music as well as automating piano playing is researched, developed and built. The project was divided into three sub-systems, the first focusing on the digitalizing of sheet music, the second on identifying and classifying the notes and the third on playing the piano. These were later combined to form a demonstrator called Mozart2000, or M2k. The result was a robot which could determine the note pitch of an arbitrary note, or note combination, written in common music notation, and furthermore play these on the piano. The algorithm is based of off finding coordinates for stafflines and notes using image processing. Programming was done in Python with some functions extracted from the library OpenCV (Open Source Computer Vision). The piano playing mechanism uses solenoids and lever arms, controlled by electrical signals from a Raspberry Pi. Due to scope in budget and time some restrictions were made. The note range for the robot was limited to one octave, meaning 8 piano keys. Moreover, other musical information such as rhythmical and coloring were overlooked and set to a predetermined value. For the digitalizing part, a camera was used, taking a snapshot of one musical bar. The final solution however can be expanded to include additional keys and music segments by replicating the existing mechanism. / I detta projekt söks en lösning för digitalisering av tryckt notskrift, mot bakgrund av en ökande efterfrågan på digitala lösningar. Parallellt undersöks möjligheterna till att automatisera pianospel. För att underlätta arbetet delades projektet in i tre delsystem; det första fokuserade på digitalisering av notpapper, det andra på att hitta och identifiera noter, och det tredje på pianospelet. Delsystemen kunde därefter integreras och resulterade då i Mozart2000, M2k. Den slutgiltiga lösningen är en robot som kan bestämma tonhöjden från ett notpapper och spela dessa på ett piano. Den framtagna algoritmen bygger på att hitta koordinater för notlinjer och noter, jämföra dessa sinsemellan och tilldela dessa utgångar på en Raspberry Pi. Från dessa skickas elektriska signaler till en krets bestående av bland annat transistorer och frihjulsdioder, som i sin tur är kopplade till solenoider. Dessa solenoider kopplade till egentillverkade fingrar kommer sedan att slå an tangenterna på pianot. Eftersom projektet var begränsat i tid och budget gjordes ett antal förenklingar. Till exempel skulle Mozart2000 hålla sig till en oktav, det vill säga åtta toner. Vidare skulle rytmen vara en konstant och endast en takt skulle analyseras och spelas åt gången. Det bedöms dock möjligt att duplicera systemet för att täcka ett större notomfång och/eller fler takter.

Nous présentons dans cette thèse une méthode complète de reconnaissance de partitions musicales imprimées, dans le cas monodique. Le système procède en deux phases distinctes : - La segmentation et l'analyse des symboles (essentiellement par corrélation), conçues pour surmonter les difficultés liées aux interconnexions et aux défauts d'impression, aboutissant à des hypothèses de reconnaissance. - L'interprétation de haut niveau, fondée sur une modélisation floue des informations extraites de l'image et des règles de notation, menant à la décision. Dans cette approche, la décision est reportée tant que le contexte n'est pas entièrement connu. Toutes les configurations d'hypothèses sont successivement évaluées, et la plus cohérente est retenue, par optimisation de tous les critères. Le formalisme utilisé, fondé sur la théorie des ensembles flous et des possibilités, permet de prendre en compte les différentes sources d'imprécision et d'incertitude, ainsi que la souplesse et la flexibilité de l'écriture musicale. Afin de gagner en fiabilité, nous proposons également des méthodes d'indication automatique des erreurs potentielles de reconnaissance, ainsi qu'une procédure d'apprentissage, optimisant les paramètres du système pour le traitement d'une partition particulière. Les performances obtenues sur une large base de données ont permis de montrer l'intérêt de la méthode proposée.

The diploma thesis specifies digital methods of optical recognition of a notation, by detailed analysis of methods based on removal of notation lines and creation of a test program which automatically converts the images written in the notation into digital format. This work summarizes the knowledge from the research and practical part. In the research section, key chapters are described as OMR architecture, including processing, symbol classification, postprocessing, and more. The practical part of the thesis presents the results of the development and testing of the proposed application.

This work deals with development of an application for optical music recognition. This application is designed for mobile phones with Android operating system. The work includes a brief introduction to the problem and introduces some existing solutions. There are described methods for image processing and classification, which are used in final application. It also shows the design and implemenation of the final aplication, where used methods for detection and removal of staff lines, detection and processing of musical symbols and their classification are described. The evaluation of the final application and a summary of achieved results are shown in the end of this work.

Building Integrated Photo Voltaics (BIPVs) are devices which are manufactured to replace building components exposed to sufficient sunlight to generate energy. Photo Voltaic Roof tiles are Building Integrated components which can be used instead of traditional roofing materials. The following thesis is focused on comparing traditional, cheaper asphalt roof tiles with Photo Voltaic (PV) roofing tiles in terms of energy cost savings during their respective Net Present Values. The method used for achieving this is computer simulation made possible by software named "Solar Advisory Model" (SAM), developed by National Renewable Energy Laboratories (NREL), to simulate energy output and resultant energy costs saved. The simulations have been run on a prototype example of a model of a dwelling unit's roof area. The simulations have been repeated for 35 cities all over the U.S.A. for 5 different climatic zones on the same prototype example of the dwelling unit. Similarly, the roof area being laid with an array of PV roof tiles has been estimated for coverage by traditional asphalt roof shingles by using data from the RS Means construction costs data. The estimated costs associated with the asphalt roof area have been adjusted to a different set of 35 locations from the 5 climatic zones by using the location factor from RS Means. A statistical analysis was done to analyze the data, net present value of roofing materials being the dependent variable versus climatic zones and roofing material as the independent variables. The statistical model also included CDD (Cooling Degree Days) and HDD (Heating Degree Days) as co-variates. The results indicate that NPV (Net Present Value) of BIPV roof is significantly different from that of asphalt roof. Another statistical analysis was done to determine the effect of climatic zones on energy savings due to the use of BIPV roofing. Energy savings (in US$) was used as a dependent variable, and climatic zone as the independent variable. HDD AND CDD were also included in this model as co-variates. The results of this test indicate that both climatic zone and HDD have an effect on total energy savings.

The purpose of this study is to investigate how five individuals ended up in a criminal lifestyle and what other factors may have had an impact on the path to a criminal lifestyle. The study is based on qualitative interviews in combination with a thematic analysis method, to find patterns around the respondents statements. As men are generally overrepresented in terms of criminal groups in crime statistics, we have chosen to limit the study to only examining men who have had a previous criminal lifestyle. The study’s theoretical framework and earlier research show that the environment and social interactions have an effect on however a person develops a criminal lifestyle. Earlier research opined that social ties to society, family, friends and school have a major impact on whether a human becomes a criminal or not. The study’s results show four prominent themes that the respondents have described in their life stories; identity, relationships, sense of community and living environment and financial vulnerability. The study concludes that all themes affect and interact with each other, which affects whether an individual develops a criminal lifestyle. The study shows that an individual ends up in a criminal lifestyle due to external factors and thus does not make an active choice to become a criminal. As this study is based on a very small sample group, further research with a larger sample group is relevant in order to better understand the reasons why an individual ends up in a criminal lifestyle.