An AdaBoost algorithm for construction of strong classifier from several weak hypotesis will be presented in this work. Theoretical background of the algorithm and the method of construction of strong classifiers will be explained. WaldBoost extension to the algorithm will be described. The thesis deals with image features that are often used as element of weak classifiers. Brief introduction to pattern recognition in context of computer vision will be outlined in the begining of the work. Also some widely used methods of classifier training will be presented. An object detection library based on AdaBoost classifiers was developed as part of the work. The library was used in implementation of software that in praktice demonstrates object detection in videosquences. Last part of the work describes tool for training of AdaBoost classifiers.

Biometric identification uses individuals' characteristics to attempt to match a sample to a database of existing samples. An increasingly commonly used characteristic is the iris section of the eye, which is valued for its uniqueness among individuals and stability over time. One key concern with iris recognition systems is the time required to find a test sample's match in a database of subjects. This work considers methods for categorizing irises within a database, so that a search for a match to a test sample can be focused on the test sample's category. The main method for categorization used in this work is texton learning. Texton learning involves creating a vocabulary of features and determining how much of each feature a given sample has. Once images are represented by textons, they are clustered in an unsupervised process. Success of the system is assessed as its ability to take a previously unseen image from a subject and classify it the same as the database reference for the subject. This work improves upon the past applications of texton learning with more thorough experiments to determine the optimal number of textons and image clusters. This system also investigates different accuracy metrics, with this work detailing two key methods and their relative benefits. Additionally, more in depth analysis is given for potential time saving impacts for finding a database match. Beyond the improvements to texton learning, symbolic features (ethnicity and gender) have been incorporated into the categorization process using a probabilistic metric. This is an innovative combination of using the numerical representation of the iris along with demographic information.

京都大学 / 0048 / 新制・課程博士 / 博士(情報学) / 甲第19417号 / 情博第596号 / 新制||情||104(附属図書館) / 32442 / 京都大学大学院情報学研究科知能情報学専攻 / (主査)教授 山本 章博, 教授 黒橋 禎夫, 教授 鹿島 久嗣, 准教授 Cuturi Marco / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

No description available.

No description available.

No description available.

Research has shown that people often fail to notice changes to visual scenes. This phenomenon is known as change blindness. This study investigated the effect of facial feature transformations on change blindness using change detection tasks involving a person as the object of change. 301 participants viewed a photo-story comprised of a few still frames. In the final frame, a selected facial feature of a character in the story was altered. Four different photo-stories were used, each utilising a different alteration. Questionnaires designed to determine whether the change was detected were administered. Results showed that changes to facial features considered to be more salient produced higher levels of change detection. A flicker test using the same images from the photo-story was administered to a further 75 participants and showed a similar pattern of results. It was concluded that in order to detect change, the changing stimuli have to be both salient and meaningful.

Readability assessment (RA) is to assign a score or a grade to a given document, which measures the degree of difficulty to read the document. RA originated in language education studies and was used to classify reading materials for language learners. Later, RA was applied to many other applications, such as aiding automatic text simplification.  This thesis is aimed at improving the way of using Transformer for RA. The motivation is the “pipeline” effect (Tenney et al., 2019) of pretrained Transformers: lexical, syntactic, and semantic features are best encoded with different layers of a Transformer model.  After a preliminary test of a basic RA model that resembles the previous works, we proposed several methods to enhance the performance: by using a Transformer layer that is not the last, by concatenating or mixing the outputs of all layers, and by using syntax-augmented Transformer layers. We examined these enhanced methods on three datasets: WeeBit, OneStopEnglish, and CommonLit.  We observed that the improvements showed a clear correlation with the dataset characteristics. On the OneStopEnglish and the CommonLit datasets, we achieved absolute improvements of 1.2% in F1 score and 0.6% in Pearson’s correlation coefficients, respectively. We also show that an 𝑛-gram frequency- based baseline, which is simple but was not reported in previous works, has superior performance on the classification datasets (WeeBit and OneStopEnglish), prompting further research on vocabulary-based lexical features for RA.

Despite the importance of pets in people’s lives and the possible emotional value of purchasing a pet, there is a lack of niche pet marketplaces. There is also a gap in research regarding the design of online marketplaces for finding pets and how users experience them. With this thesis project, I aim to touch upon that gap and provide insights to build upon in future research. This is achieved by providing recommendations of features that can enhance the user experience by addressing existing needs and concerns. To answer the research questions, interviews and a workshop were conducted followed by an analysis. Based on the findings, prototypes were designed and evaluated. Finally, the results of the study are discussed from a user experience perspective. / Trots betydelsen som husdjur kan ha i människors liv och det möjliga känslomässiga värdet av att köpa ett husdjur så finns det brist på nischade marknadsplatser där man kan köpa djur. Det saknas även forskning kring utformningen av digitala marknadsplatser för djur och hur användare upplever dessa. Med denna studie strävar jag efter att beröra detta ämne och bidra med insikter att bygga vidare på i framtida forskning. Detta uppnås genom att tillhandahålla rekommendationer om funktioner som kan förbättra användarupplevelsen genom att ta hänsyn till befintliga behov och problem. För att svara på forskningsfrågorna genomfördes intervjuer och en workshop följt av en analys. Baserat på resultaten utvecklades prototyper som sedan även utvärderades. Slutligen diskuteras studiens resultat ur perspektivet för användarupplevelse.

Introduction: Borderline personality disorder (BPD) is a severe psychiatric disorder characterized by emotion dysregulation, interpersonal dysfunction, and poor impulse control. Little research has investigated BPD in the context of major life events. The perinatal period (pregnancy until 12 months postpartum) is an important milestone that involves major role transitions and novel challenges. This thesis examined the associations between borderline personality features (BPF) and depressive and generalized anxiety symptoms in a sample of treatment-seeking perinatal women. Methods: 74 perinatal women were recruited from the Women’s Health Concerns Clinic (WHCC) at St. Joseph’s Healthcare Hamilton, Canada, and enrolled in the WHCC Registry study. Participants were sent online intake questionnaires to collect data about demographic, personality, and other psychosocial variables. They also completed three self-report mental health measures: the McLean Screening Instrument for Borderline Personality Disorder (MSI-BPD), the Edinburgh Postnatal Depression Scale (EPDS), and the Generalized Anxiety Disorder Scale (GAD-7). Logistic regression was used to determine whether a positive MSI-BPD screen (score ≥ 7) was associated with a higher likelihood of screening positive on the EPDS or GAD-7 (score ≥ 13). Results: A positive screen on the MSI-BPD was significantly associated with an almost eighteen-fold increase in the odds of screening positive on the EPDS in our treatment-seeking perinatal sample (OR 17.84, 95% CI[2.11, 218.80], p<0.05). A positive screen on the MSI-BPD was not associated with higher odds of screening positive on the GAD-7, rather only childhood trauma and a positive screen on the EPDS emerged as significant predictor variables. Our findings may reflect the greater symptomatic overlap observed between BPD and perinatal depression as well as the comparatively lower comorbidity observed between GAD and BPD in non-perinatal research. Conclusions and Future Directions: The use of self-report measures, low statistical power, and a treatment-seeking sample are limitations to consider when interpreting our findings. To our knowledge, this research study offers one of the first explorations into the relationship between BPD and generalized anxiety symptoms during the perinatal period. Future research should aim to better characterize perinatal BPD and investigate its relationship with other mental health conditions. / Thesis / Master of Science (MSc) / The perinatal period (pregnancy until 12 months postpartum) represents a time of heightened vulnerability to poor mental health. Prior research has mainly focused on perinatal depression and anxiety, while perinatal personality disorders have received comparably less attention. Borderline personality disorder (BPD) is a severe psychiatric disorder associated with diminished ability to regulate emotions, disturbances in self-image, troubled interpersonal relationships, and impulsive behaviour. This thesis investigated the relationship between self-reported borderline personality features (BPF) and depressive and generalized anxiety symptoms in a sample of perinatal women seeking treatment at a psychiatric clinic. We hope that this research sheds light on the nature of perinatal BPD, as well as its associations with other mental health conditions, to improve both immediate and multi-generational maternal and infant well-being.