Avec le développement des appareils photos numériques et des sites de partage de photos, nous passons une part croissante de notre temps à observer, sélectionner et partager des images, parmi lesquelles figurent un grand nombre de photos de visage. Dans cette thèse, nous nous proposons de créer un premier système entièrement automatique renvoyant une estimation de la pertinence d'une photo de visage pour son utilisation dans la création d'un album de photos, la sélection de photos pour un réseau social ou professionnel, etc. Pour cela, nous créons plusieurs modèles d'estimation de la pertinence d'une photo de visage en fonction de son utilisation. Dans un premier temps, nous adaptons les modèles d'estimation de la qualité esthétique d'une photo au cas particulier des photos de visage. Nous montrons que le fait de calculer 15 caractéristiques décrivant différents aspects de l'image (texture, illumination, couleurs) dans des régions spécifiques de l'image (le visage, les yeux, la bouche) améliore significativement la précision des estimations par rapport aux modèles de l'état de l'art. La précision de ce modèle est renforcée par la sélection de caractéristiques adaptées à notre problème, ainsi que par la fusion des prédictions de 4 algorithmes d'apprentissage. Dans un second temps, nous proposons d'enrichir l'évaluation automatique d'une photo de visage en définissant des modèles d'estimation associés à des critères tels que le degré de sympathie ou de compétence dégagé par une photo de visage. Ces modèles reposent sur l'utilisation d'attributs de haut niveau (présence de sourire, ouverture des yeux, expressions faciales), qui se montrent plus efficaces que les caractéristiques de bas niveau utilisées dans l'état de l'art (filtres de Gabor, position des points de repère du visage). Enfin, nous fusionnons ces modèles afin de sélectionner automatiquement des photos de bonne qualité esthétique et appropriées à une utilisation donnée : photos inspirant de la sympathie à partager en famille, photos dégageant une impression de compétence sur un réseau professionnel. / Picture selection is a time-consuming task for humans and a real challenge for machines, which have to retrieve complex and subjective information from image pixels. An automated system that infers human feelings from digital portraits would be of great help for profile picture selection, photo album creation or photo editing. In this work, several models of facial pictures evaluation are defined. The first one predicts the overall aesthetic quality of a facial image by computing 15 features that encode low-level statistics in different image regions (face, eyes and mouth). Relevant features are automatically selected by a feature ranking technique, and the outputs of 4 learning algorithms are fused in order to make a robust and accurate prediction of the image quality. Results are compared with recent works and the proposed algorithm obtains the best performance. The same pipeline is then considered to evaluate the likability and competence induced by a facial picture, with the difference that the estimation is based on high-level attributes such as gender, age and smile. Performance of these attributes is compared with previous techniques that mostly rely on facial keypoints positions, and it is shown that it is possible to obtain predictions that are close to human perception. Finally, a combination of both models that selects a likable facial image of good aesthetic quality for a given person is described.

For decades, cement mortar relining has been used successfully to extend the life of drinking water pipelines, although, few quantitative data exist on the short-term water quality impacts of this process. This study investigated mortar lining impacts on disinfectant by-product formation, alkalinity, metal leaching, pH and disinfectant consumption, and odor generation shortly after in-situ installation. The experimental design used a 30-day, coupon immersion procedure that simulated a relined 4-inch diameter pipe located in a low-flow system. Four water regimes were utilized; no disinfectant, chlorine (2 mg/L at pH 6 .5 and 8), and chloramines. Flavor Profile Analysis panels evaluated odors of samples and controls. Additionally, the affects of three different phosphate-based corrosion prevention additive regimes were evaluated. Cement mortar leachates impacted water quality significantly during the first week of exposure. Alkalinity, hardness and pH increased dramatically after initial exposure, rising to approximately 600 mg/L as CaCO3 alkalinity, 770 mg/L as CaCO3 hardness, and pH 12 in the first two days. Sharp declines in alkalinity and hardness did not occur until after day 9 when the cement mortar was substantially cured and release of calcium hydroxide lessened. Chlorinated water residual disinfectant decay rate was increased substantially during the initial 24 hours and remained elevated until day 9. After day 1, there was not a significant increase in the formation of regulated haloacetic acids or trihalomethanes. Significant levels of aluminum (< 700 ug/L) and chromium (< 75 ug/L) were released at various times during the test period but their concentrations did not exceed USEPA water quality standards. Cement odor intensity levels were significantly higher than controls, persisted for 14 days, and were of an intensity that would be readily noticeable to consumers. The polyphosphate-based corrosion preventative resulted in less severe water quality effects than other phosphate additives or water without added phosphate. / Master of Science

Consumer satisfaction is key for retailers to retain loyal consumers. Retailers use apparel brands as a strategy to differentiate themselves in the market place, to attract potential consumers and to retain loyal consumers. Retailers have discovered the potential of the fuller figure sector as a lucrative market which can assist them to make profits and to increase their sales. The aim of this research is firstly to explore and describe the needs and expectations that the fuller figure South African working women have with branded career wear secondly, how they evaluate branded career wear apparel against the two broad quality dimensions namely functional and aesthetic dimensions and thirdly the problems they experience with the functional and aesthetic quality dimensions of branded career wear. The study also describes how fuller figure South African working women consumers further evaluate the branded apparel during the decision-making process which can ultimately lead to her either accepting or rejecting the branded career wear. This research is descriptive in nature, as an attempt is made to describe and understand how working fuller figure females evaluate branded apparel against certain quality dimensions and make decisions to purchase branded apparel to satisfy their needs. The Sproles and Burns’ (1994:264) decision-making model theoretical underpins the study. A quantitative research style was chosen for this study and the objectives and sub- objectives related to the needs and problems of the quality dimensions of the branded career wear were examined. A self-administered, structured questionnaire was designed to gain demographic information of the sample group, their store preferences, where they get their fashion information from as well as their needs and problems with regard to branded career wear apparel. Purposive sampling and snowball sampling techniques were used to select a sample of working fuller figure females in the Tshwane, South Africa area. The responses to the 150 questionnaires were coded, captured and analysed using descriptive and inferential statistics. The results of this study revealed that most of the respondents ranked the functional quality dimensions higher than the aesthetic quality dimensions. It is clear that most of the fuller figure consumers regard comfort as the most important need, followed by durability (that the material of the branded career wear should be of good quality). The sample group also regarded the emotional aesthetic aspects very important (that the styles are beautiful and make her feel feminine). It is contradictory, that these fuller figure working women experienced major problems with the quality dimensions that they ranked as highly important. The sample group experienced problems with fit at the functional and aesthetic level, although fit is one of the most important factors that consumers use during the evaluation of apparel items. This study makes certain recommendations to apparel retailers, fashion theorists and image consultants how to obtain optimal consumer satisfaction. The South African apparel retailers should take cognisance of the needs and problems of this growing market related to the functional and aesthetic quality dimensions. The study also revealed how age and income play a role regarding the needs and problems this sample group experienced with branded career wear and how this can influence their decision to either accept or reject a brand. The results contribute to the body of knowledge regarding the needs and problems of branded career wear apparel for the fuller figure market related to certain functional and aesthetic quality dimensions Copyright / Dissertation (MConsumer Science)--University of Pretoria, 2012. / Consumer Science / unrestricted