Le cancer du sein est le type de cancer le plus répandu, il est la cause principale de mortalité chez les femmes aussi bien dans les pays occidentaux que dans les pays en voie de développement. L'imagerie médicale joue un rôle clef dans la réduction de la mortalité du cancer du sein, en facilitant sa première détection par le dépistage, le diagnostic et la biopsie guidée. Bien que la Mammographie Numérique (DM) reste la référence pour les méthodes d'examen existantes, les échographies ont prouvé leur place en tant que modalité complémentaire. Les images de cette dernière fournissent des informations permettant de différencier le caratère bénin ou malin des lésions solides, ce qui ne peut être détecté par DM. Malgré leur utilité clinique, les images échographiques sont bruitées, ce qui compromet les diagnostiques des radiologues à partir de celles ci. C'est pourquoi un des objectifs premiers des chercheurs en imagerie médicale est d'améliorer la qualité des images et des méthodologies afin de simplifier et de systématiser la lecture et l'interprétation de ces images.La méthode proposée considère le processus de segmentation comme la minimisation d'une structure probabilistique multi-label utilisant un algorithme de minimisation du Max-Flow/Min-Cut pour associer le label adéquat parmi un ensemble de labels figurant des types de tissus, et ce, pour tout les pixels de l'image.Cette dernière est divisée en régions adjacentes afin que tous les pixels d'une même régions soient labelisés de la même manière en fin du processus. Des modèles stochastiques pour la labellisation sont crées à partir d'une base d'apprentissage de données. / Breast cancer is the second most common type of cancer being the leading cause of cancer death among females both in western and in economically developing countries. Medical imaging is key for early detection, diagnosis and treatment follow-up. Despite Digital Mammography (DM) remains the reference imaging modality, Ultra-Sound (US) imaging has proven to be a successful adjunct image modality for breast cancer screening, specially as a consequence of the discriminative capabilities that US offers for differentiating between solid lesions that are benign or malignant. Despite US usability,US suffers inconveniences due to its natural noise that compromises the diagnosis capabilities of radiologists. Therefore the research interest in providing radiologists with Computer Aided Diagnosis (CAD) tools to assist the doctors during decision taking. This thesis analyzes the current strategies to segment breast lesions in US data in order to infer meaningful information to be feet to CAD, and proposes a fully automatic methodology for generating accurate segmentations of breast lesions in US data with low false positive rates.

This work deals with the design and implementation of a web application for playback and annotation of multimedia data from files in BAG format. The created annotations are intended for training artificial intelligence in robotic systems. The resulting application is implemented with the help of React library for creating the user interface and framework Django for the server-side REST API. The entire application is containerized using Docker. This work describes the BAG format, analyzes modern libraries for creating web applications and introduces existing solutions for data annotation or playback of BAG files. Furthermore, the design of the application is presented and its implementation is described. A separate library rosbag_pyreader, which is used to read BAG files with random access, was created as part of this work. The resulting application allows user to display and annotate 2D and 3D data from multiple robotic sensors at the same time and then export the annotations in JSON format.

Submitted by Angela Maria de Oliveira (amolivei@uepg.br) on 2018-11-20T18:16:09Z No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) Flavia de Oliveira.pdf: 1587499 bytes, checksum: 2fb58a43f13e11f2b66666d67ce47b88 (MD5) / Made available in DSpace on 2018-11-20T18:16:09Z (GMT). No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) Flavia de Oliveira.pdf: 1587499 bytes, checksum: 2fb58a43f13e11f2b66666d67ce47b88 (MD5) Previous issue date: 2018-08-31 / Sistemas integrados de produção agropecuária (SIPA) integram em um mesmo espaço vários componentes, tais como lavoura, pastagens, animais, bem como árvores de forma sinérgica. Quando bem manejado, pode trazer a sustentabilidade para a propriedade. Um dos desafios em tais sistemas é o entendimento da dinâmica de nutrientes e o possível reaproveitamento destes pela cultura sucessora. O objetivo deste trabalho foi avaliar o aporte de liteira e dos macronutrientes, como nitrogênio (N), fósforo (P), potássio (K) e enxofre (S) em SIPA, com ausência e presença de árvores (~159 árvores ha-1), submetidos a doses de nitrogênio (90 e 180 kg N ha-1, N90 e N180, respectivamente) na fase pastagem. O experimento foi conduzido no Instituto Agronômico do Paraná, município de Ponta Grossa, Paraná. O delineamento experimental utilizado foi o de blocos ao acaso com quatro tratamentos e três repetições. De modo a melhor representar o ambiente arborizado, foi feita a subdivisão da unidade experimental em três posições entre linha das árvores. O experimento iniciou em 2006, com o plantio de árvores em arranjo de 14 x 3m. Durante o inverno cultiva-se aveia preta e azevém, pastejada por bovinos de corte, e, durante o verão, soja/milho. O presente experimento iniciou na fase pastagem de 2013, a qual antecedeu a cultura do milho (safra 2013/2014), estendendo-se até a fase pastagem no inverno de 2014. Durante a lavoura de milho e pastagem (2014), foram inseridos litter bags com os resíduos da cultura anterior, no momento das suas respectivas semeaduras. Aos 8, 15, 30, 60, 90, 120, 153 e 165 dias após a semeadura do milho e da pastagem foram coletadas as amostras dos litter bags, pesadas suas respectivas massas secas (MS) e quantificados os macronutrientes. Também foram calculadas as taxas de decomposição do resíduo e da liberação de macronutrientes. Em geral, a quantidade inicial de resíduo, tanto da pastagem como do milho, foi inferior nos tratamentos com árvores (-28%). Diferenças significativas foram observadas entre os tratamentos nos teores iniciais em N (variando de 12 no ILPF N90 até 14 g/kg no ILPF N180) do resíduo da pastagem e para os teores iniciais de K (variando entre 6,5 no ILP N90 até 12 g/kg no ILPF N90) e S (variando de 0,68 ILPF N90 até 0,91g/kg ILP N90) no resíduo do milho. As posições entre as linhas das árvores afetaram apenas os teores iniciais de N e S dos resíduos. Em relação a dinâmica de decomposição dos resíduos e liberação de macronutrientes, diferenças significativas entre os tratamentos foram observadas somente para decomposição do resíduo da pastagem e liberação de P do mesmo. A fração ativa da MS da pastagem variou entre 44 (ILPF N90) a 59 % (ILP N180). O tempo de meia vida (T1/2) variou entre 11 (ILP N90) e 54 (ILPF N180) dias para MS e entre 6 (ILP N90) e 21 (ILPF N90) dias para o P. O T1/2 do N foi de 19 dias, enquanto que para K e S foi de 13 dias. Em relação ao resíduo do milho, o T1/2 variou entre 5 (no caso de N) e 24 (no caso de P) dias. No mínimo 44% do resíduo proveniente da pastagem foi decomposto ao longo dos 165 dias de exposição à campo dos litter bags. Em relação ao resíduo do milho, 30% foi decomposto ao longo de 90 dias, independente do tratamento. Máximas liberações (37; 2,8; 62 e 2,5 kg/há de N, P, K e S, respectivamente) de nutrientes foram observadas no ILP N180 para o resíduo da pastagem, principalmente em função da maior quantidade de resíduo inicial neste tratamento. Para o resíduo do milho, máximas liberações ocorreram no ILP N90 para N (48 kg/ha), P (17,4 kg/ha) e S (4,8 kg/ha), em função da maior quantidade de resíduo inicial neste tratamento, e no ILPF N90 para K (82 kg/ha), em função do maior teor inicial em K. Portanto, tais quantidades cicladas devem ser consideradas no manejo da fertilização. Desbastes precisam ser intensificados no ILPF para reduzir o nível de sombreamento e evitar grandes reduções nas quantidades de resíduo, para otimizar os benefícios da ciclagem de nutrientes. / Integrated Crop Livestock Systems (ICLS) comprehend in a same space several components, such as grain and forages crops, animals and tree component of synergistic form. When well managed, it can bring sustainability to the property. One of the challenges in such systems is the understanding of the nutrient dynamics and the possible reutilization of these by the successor culture. The objective of this study was to evaluate the contribution of litter and macronutrients, such as nitrogen (N), phosphorus (P), potassium (K) and sulfur (S) in ICLS, with absence and presence of trees ), submitted to nitrogen doses (90 and 180 kg N ha-1, N90 and N180, respectively) in the pasture phase. The experiment was carried out at the Agronomic Institute of Paraná, municipality of Ponta Grossa, Paraná. The experimental design was a randomized block design with four treatments (ILP N90, ILP N180, ILPF N90 and ILPF N180, being ILP, integration of livestock and ILPF, integration of livestock and forest) and three replications. In order to better represents the wooded environment, the experimental unit was subdivided into three positions (1m, 3.5m and 7m distance from the tree line). The experiment began in 2006 with the planting of trees in arrangement of 14 x 3m. During winter, black oats and ryegrass are grazed by beef cattle and, in the summer, soybean/corn. The present experiment started in the pasture phase of 2013, which preceded the corn crop (crop 2013/2014), extending until the pasture stage in the winter of 2014. During maize and pasture (2014), litter bags were inserted with the residues from the previous crop, at the time of their respective sowing. The samples were collected at 8, 15, 30, 60, 90, 120, 153, and 165 days after sowing of maize and pasture. The samples were performed to dry masses (DM) measure and macronutrients determination. The rates of decomposition of the residue and the release of macronutrients were also calculated. Overall, the initial amount of residue from both pasture and maize was lower in tree treatments (-28%). Differences were observed between treatments at initial N contents (ranging from 12 in ILPF N90 to 14 g / kg in ILPF N180) of the pasture residue and for the initial K contents (ranging from 6.5 in ILP N90 to 12 g / kg in ILPF N90) and S (ranging from 0.68 ILPF N90 to 0.91 g / kg ILP N90) in the corn residue. The positions between the tree lines affected only the initial N and S contents of the residues. In relation to the dynamics of the decomposition of the residues and the release of macronutrients, significant differences between the treatments were observed only for the decomposition of the pasture residue and the release of P from the same. The active fraction of DM from pasture ranged from 44 (ILPF N90) to 59% (ILP N180). The half-life time (T1/2) varied between 11 (ILP N90) and 54 (ILPF N180) days for MS and between 6 (ILP N90) and 21 (ILPF N90) days for P. O T1/2 of N was 19 days, whereas for K and S it was 13 days. In relation to maize residue, T1/2 ranged from 5 (in the case of N) to 24 (in the case of P) days. At least 44% of the pasture residue was decomposed during the 165 days of exposure to litter bags. Regarding corn residue, 30% was decomposed over 90 days, regardless of the treatment. The maximum nutrient release (37, 2.8, 62 and 2.5 kg/ha of N, P, K and S respectively) was observed in the ILP N180 for the pasture residue, mainly due to the higher amount of initial residue in this treatment. For the maize residue, maximum releases occurred in ILP N90 for N (48 kg/ha), P (17.4 kg/ha) and S (4.8 kg / ha), due to the higher amount of initial residue in this treatment, and in ILPF N90 for K (82 kg / ha), as a function of the higher starting content in K. Therefore, such cycled quantities should be considered in the management of fertilization. Slabs need to be intensified in the ILPF to reduce the level of shading and avoid large reductions in residue amounts to optimize the benefits of nutrient cycling.

Les véhicules automobiles autonomes sont des systèmes critiques. En effet, suite à leurs défaillances, ils peuvent provoquer des dégâts catastrophiques sur l'humain et sur l'environnement dans lequel ils opèrent. Le contrôle des véhicules autonomes robotisés est une fonction complexe, qui comporte de très nombreux modes de défaillances potentiels. Dans le cas de plateformes expérimentales qui n'ont suivi ni les méthodes de développement ni le cycle de certification requis pour les systèmes industriels, les probabilités de défaillances sont beaucoup plus importantes. En effet, ces véhicules expérimentaux se heurtent à deux problèmes qui entravent leur sûreté de fonctionnement, c'est-à-dire la confiance justifiée que l'on peut avoir dans leur comportement correct. Tout d'abord, ils sont utilisés dans des environnements ouverts, au contexte d'exécution très large. Ceci rend leur validation très complexe, puisque de nombreuses heures de test seraient nécessaires, sans garantie que toutes les fautes du système soient détectées puis corrigées. De plus, leur comportement est souvent très difficile à prédire ou à modéliser. Cela peut être dû à l'utilisation des logiciels d'intelligence artificielle pour résoudre des problèmes complexes comme la navigation ou la perception, mais aussi à la multiplicité de systèmes ou composants interagissant et compliquant le comportement du système final, par exemple en générant des comportements émergents. Une technique permettant d'augmenter la sécurité-innocuité (safety) de ces systèmes autonomes est la mise en place d'un composant indépendant de sécurité, appelé « Safety-Bag ». Ce système est intégré entre l'application de contrôle-commande et les actionneurs du véhicule, ce qui lui permet de vérifier en ligne un ensemble de nécessités de sécurité, qui sont des propriétés nécessaires pour assurer la sécurité-innocuité du système. Chaque nécessité de sécurité est composée d'une condition de déclenchement et d'une intervention de sécurité appliquée quand la condition de déclenchement est violée. Cette intervention consiste soit en une inhibition de sécurité qui empêche le système d'évoluer vers un état à risques, soit en une action de sécurité afin de remettre le véhicule autonome dans un état sûr. La définition des nécessités de sécurité doit suivre une méthode rigoureuse pour être systématique. Pour ce faire, nous avons réalisé dans nos travaux une étude de sûreté de fonctionnement basée sur deux méthodes de prévision des fautes : AMDEC (Analyse des Modes de Défaillances, leurs Effets et leur Criticité) et HazOp-UML (Etude de dangers et d'opérabilité) qui mettent l'accent respectivement sur les composants internes matériels et logiciels du système et sur l'environnement routier et le processus de conduite. Le résultat de ces analyses de risques est un ensemble d'exigences de sécurité. Une partie de ces exigences de sécurité peut être traduite en nécessités de sécurité implémentables et vérifiables par le Safety-Bag. D'autres ne le peuvent pas pour que le système Safety-Bag reste un composant relativement simple et validable. Ensuite, nous avons effectué des expérimentations basées sur l'injection de fautes afin de valider certaines nécessités de sécurité et évaluer le comportement de notre Safety-Bag. Ces expériences ont été faites sur notre véhicule robotisé de type Fluence dans notre laboratoire dans deux cadres différents, sur la piste réelle SEVILLE dans un premier temps et ensuite sur la piste virtuelle simulée par le logiciel Scanner Studio sur le banc VILAD. Le Safety-Bag reste une solution prometteuse mais partielle pour des véhicules autonomes industriels. Par contre, il répond à l'essentiel des besoins pour assurer la sécurité-innocuité des véhicules autonomes expérimentaux. / Autonomous automotive vehicles are critical systems. Indeed, following their failures, they can cause catastrophic damage to the human and the environment in which they operate. The control of autonomous vehicles is a complex function, with many potential failure modes. In the case of experimental platforms that have not followed either the development methods or the certification cycle required for industrial systems, the probabilities of failure are much greater. Indeed, these experimental vehicles face two problems that impede their dependability, which is the justified confidence that can be had in their correct behavior. First, they are used in open environment, with a very wide execution context. This makes their validation very complex, since many hours of testing would be necessary, with no guarantee that all faults in the system are detected and corrected. In addition, their behavior is often very difficult to predict or model. This may be due to the use of artificial intelligence software to solve complex problems such as navigation or perception, but also to the multiplicity of systems or components interacting and complicating the behavior of the final system, for example by generating behaviors emerging. A technique to increase the safety of these autonomous systems is the establishment of an Independent Safety Component, called "Safety-Bag". This system is integrated between the control application and the actuators of the vehicle, which allows it to check online a set of safety necessities, which are necessary properties to ensure the safety of the system. Each safety necessity is composed of a safety trigger condition and a safety intervention applied when the safety trigger condition is violated. This intervention consists of either a safety inhibition that prevents the system from moving to a risk state, or a safety action to return the autonomous vehicle to a safe state. The definition of safety necessities must follow a rigorous method to be systematic. To do this, we carried out in our work a study of dependability based on two fault prevention methods: FMEA and HazOp-UML, that respectively focus on the internal hardware and software components of the system and on the road environment and driving process. The result of these risk analyzes is a set of safety requirements. Some of these safety requirements can be translated into safety necessities, implementable and verifiable by the Safety-Bag. Others cannot be implemented in the Safety-Bag. The latter must remain simple so that it is easy to be validated. Then, we carried out experiments based on the faults injection in order to validate some safety necessities and to evaluate the Safety-Bag's behavior. These experiments were done on our robotic vehicle type Fluence in our laboratory in two different settings, on the actual track SEVILLE at first and then on the virtual track simulated by the Scanner Studio software on the VILAD testbed. The Safety-Bag remains a promising but partial solution for autonomous industrial vehicles. On the other hand, it meets the essential needs for the safety of experimental autonomous vehicles.

<p>Danas, servisi e-Uprave u različitim oblastima koriste question answer sisteme koncepta u poku&scaron;aju da se razume tekst i da pomognu građanima u dobijanju odgovora na svoje upite u bilo koje vreme i veoma brzo. Automatsko mapiranje relevantnih dokumenata se ističe kao važna aplikacija za automatsku strategiju klasifikacije: upit-dokumenta. Ova doktorska disertacija ima za cilj doprinos u identifikaciji nestruktuiranih dokumenata i predstavlja važan korak ka razja&scaron;njavanju uloge eksplicitnih koncepata u pronalaženju podataka uop&scaron;te ajče&scaron; a reprezenta vna &scaron;ema u tekstualnoj kategorizaciji je BoW pristup, kada je u pozadini veliki skup znanja. Ova disertacija uvodi novi pristup ka stvaranju koncepta zasnovanog na tekstualnoj prezantaciji i primeni kategorizacije teksta, kako bi se stvorile definisane klase u slučaju sažetih tekstualnih dokumenata Takođe, ovde je prikazan algoritam zasnovan na klasifikaciji, modelovan za upite koji odgovaraju temi. Otežavaju a okolnost u slučaju ovog koncepta, koji prezentuje termine sa visokom frekvencijom pojavljivanja u upitma, zasniva se na sličnostima u prethodno definisanim klasama dokumenata Rezultati eksperimenta iz oblasti Krivičnog zakonika Republike Srbije, u ovom slučaju i studija, pokazuju da prezentacija teksta zasnovana na konceptu ima zadovoljavaju e rezultate i u slučaju kada ne postoji rečnik za datu oblast.</p> / <p>Nowadays, the concept of Question Answering Systems (QAS) has been used by e-government services in various fields as an attempt to understand the text and help citizens in getting answers to their questions promptly and at any time. Automatic mapping of relevant documents stands out as an important application for automatic classification strategy: query-document. This doctoral thesis aims to contribute to identification of unstructured documents and represents an important step towards clarifying the role of explicit concepts within Information Retrieval in general. The most common scheme in text categorization is BoW approach, especially when, as a basis, we have a large set of knowledge. This thesis introduces a new approach to the creation of text presentation based concept and applying text categorization, with the aim to create a defined class in case of compressed text documents.Also, this paper discusses the classification based algorithm modeled for queries that suit the theme. What makes the situation more complicated is the fact that this concept is based on the similarities in previously defined classes of documents and terms with a high frequency of appearance presented in queries. The results of the experiment in the field of the Criminal Code, and this paper as well, show that the text presentation based concept has satisfactory results even in case where there is no vocabulary for certain field.</p>

Ständig förändring är strakt förknippat med modeindustrin, då nya trender och stilar tenderar att förändras kontinuerligt. I västvärlden konsumerar människor betydligt mer modeprodukter än vad naturens resurser tillåter. Modekonsumenterna fick dock en insyn av textilbranschens negativa miljöpåverkan år 2017. Det skrevs då en ny lag som föreskriver att modeföretag måste informera sina modekonsumenter om bärkassarnas, speciellt plastbärkassens, negativa påverkan. I samband med att lagen stiftades, skapades organisationen One Bag Habit som fick modekonsumenterna att tänka en extra gång innan de konsumerar bärkassar i modebutikerna. På grund av bland annat One Bag Habit avstår modekonsumenter numera att konsumera bärkassar på grund av miljöaspekter, dock fortsätter de att konsumera mängder av modeprodukter utan att blicka. Modekonsumenter har sedan år 2017 blivit medvetna om de problem som bärkassen medför och väljer då att avstå dessa, men när det kommer till modeprodukter finns troligtvis en medvetenhet om problemet där också men konsumtionen fortsätter ändå. Detta har lett till att denna uppsats belyser hur denna kontrast bildas mellan att konsumera modeprodukter och att avstå från bärkassen på grund av miljöskäl. Denna studie undersöker detta gap genom att fokusera på hur modekonsumenter förhåller sig till konsumtion av bärkassar i förhållande till modeprodukter. Resultatet ger en förståelse för varför det blir en kontrast mellan konsumtion av bärkassar och modeprodukter. Därav kommer resultatet resultera i en djupare förståelse kring varför ett attityd-beteende gap skapas och en stadig grund för vidare forskning. För att förstå och tolka det empiriska materialet modifierades en modell av attityd-beteende gapet. Utöver det kommer behovsidentifikation att ligga till grund för den teoretiska referensramen. En kvalitativ intervju, tre fokusgruppsintervjuer och en observation utgör det empiriska materialet. Urvalet bestod av studenter från tre olika högskolor i Sverige (Textilhögskolan i Borås, Göteborgs Universitet och Mälardalens högskola i Västerås). Slutsatserna av studien är att modeföretagen måste börja informera modekonsumenterna varför och hur de kan agera hållbart, istället för att informera att de ska agera hållbart. Utöver det måste de hållbara modeprodukterna vara prismässigt jämförbart med inte hållbara modeprodukter, för att modekonsumenterna ska agera hållbar. Slutligen krävs det att de hållbara modeprodukterna har tillräckligt attraktiv design för att modekonsumenterna ska investera i hållbara modeprodukter. / Constant change is closely associated with the fashion industry, as new trends and styles tent to change continuously. In the Western world, people consume considerably more fashion products than what the natural resources allow. The fashion consumers, however, gained an insight into the fashion industry´s negative environmental impact in 2017. A new law was founded to alert the fashion consumers of the negative effect of the shopping bags, especially made with plastic. In association with the new law, an organization, naming One Bag Habit was created. This led the fashion consumers to think extra before deciding to consume a shopping bag when they made a purchase of fashion item. Due to One Bag habit, fashion consumers now avoid from consuming shopping bags because of environmental aspects. The reluctance to consuming a shopping bag in addition to the fashion purchase has decreased yet impacts on the actual fashion consumption have not changed. Previous research has identified an attitude-behaviour gap between how fashion consumers value sustainable products and their actual behaviour in accordance with these values. This has led to this paper highlighting how this contrast is formed between consuming fashion products and refraining from the shopping bags because of environmental reasons. This study has a textile management approach, which means that the study is primarily aimed at companies in order to help them understand how fashion consumers value and act in different purchasing situations. This study will examine how fashion consumers consume shopping bags in relation to fashion products. The findings will contribute to creating an understanding of the created contrast between sustainable consumption regarding shopping bags and non-sustainable one regarding fashion consumption itself. Moreover, the findings will contribute to a deeper understanding of why the attitude-behaviour gap has created such a contrast among the fashion consumer. In order to understand and interpret the empirical material of this study, a theoretical model of attitude-behaviour gap was created. In addition, an understanding of how needs are created among consumers will be contributing to the formation of the theoretical framework. Furthermore, data was collected through qualitative interviews, three focus groups, and an observation. The data sample frame consisted of students from three different universities in Sweden (the Swedish School of Textiles in Borås, University of Gothenburg and Mälardalens University in Västerås). The findings of the study show that fashion companies need to be more specific when informing the fashion consumers on why and how they can act more sustainably when consuming fashion and shopping bags. In addition, the sustainable fashion products should be comparable in prices terms to non-sustainable fashion products in order for fashion consumers to act sustainable. Finally, it is suggested that the sustainable fashion products have an appealing designs in order for the fashion consumers to invest in sustainable fashion products.

Computer Vision researchers are constantly challenged with questions that are motivated by real applications. One of these questions is whether a computer program could distinguish groups of people based on their geographical ancestry, using only frontal images of their faces. The advances in this research area in the last ten years show that the answer to that question is affirmative. Several papers address this problem by applying methods such as Local Binary Patterns (LBP), raw pixel values, Principal or Independent Component Analysis (PCA/ICA), Gabor filters, Biologically Inspired Features (BIF), and more recently, Convolution Neural Networks (CNN). In this work we propose to combine the Bag-of-Visual-Words model with new dictionary learning techniques and a new spatial structure approach for image features. An extensive set of experiments has been performed using two of the largest face image databases available (MORPH-II and FERET), reaching very competitive results for gender and ethnicity recognition, while using a considerable small set of images for training. / Pesquisadores de visão computacional são constantemente desafiados com perguntas motivadas por aplicações reais. Uma dessas questões é se um programa de computador poderia distinguir grupos de pessoas com base em sua ascendência geográfica, usando apenas imagens frontais de seus rostos. Os avanços nesta área de pesquisa nos últimos dez anos mostram que a resposta a essa pergunta é afirmativa. Vários artigos abordam esse problema aplicando métodos como Padrões Binários Locais (LBP), valores de pixels brutos, Análise de Componentes Principais ou Independentes (PCA/ICA), filtros de Gabor, Características Biologicamente Inspiradas (BIF) e, mais recentemente, Redes Neurais Convolucionais (CNN). Neste trabalho propomos combinar o modelo \"bag-of-words\" visual com novas técnicas de aprendizagem por dicionário e uma nova abordagem de estrutura espacial para características da imagem. Um extenso conjunto de experimentos foi realizado usando dois dos maiores bancos de dados de imagens faciais disponíveis (MORPH-II e FERET), alcançando resultados muito competitivos para reconhecimento de gênero e etnia, ao passo que utiliza um conjunto consideravelmente pequeno de imagens para treinamento.

Today, with digitally stored information available in abundance, even for many minor languages, this information must by some means be filtered and extracted in order to avoid drowning in it. Automatic summarization is one such technique, where a computer summarizes a longer text to a shorter non-rendundant form. Apart from the major languages of the world there are a lot of languages for which large bodies of data aimed at language technology research to a high degree are lacking. There might also not be resources available to develop such bodies of data, since it is usually time consuming and requires substantial manual labor, hence being expensive. Nevertheless, there will still be a need for automatic text summarization for these languages in order to subdue this constantly increasing amount of electronically produced text. This thesis thus sets the focus on automatic summarization of text and the evaluation of summaries using as few human resources as possible. The resources that are used should to as high extent as possible be already existing, not specifically aimed at summarization or evaluation of summaries and, preferably, created as part of natural literary processes. Moreover, the summarization systems should be able to be easily assembled using only a small set of basic language processing tools, again, not specifically aimed at summarization/evaluation. The summarization system should thus be near language independent as to be quickly ported between different natural languages. The research put forth in this thesis mainly concerns three computerized systems, one for near language independent summarization – The HolSum summarizer; one for the collection of large-scale corpora – The KTH News Corpus; and one for summarization evaluation – The KTH eXtract Corpus. These three systems represent three different aspects of transferring the proposed summarization method to a new language. One aspect is the actual summarization method and how it relates to the highly irregular nature of human language and to the difference in traits among language groups. This aspect is discussed in detail in Chapter 3. This chapter also presents the notion of “holistic summarization”, an approach to self-evaluative summarization that weighs the fitness of the summary as a whole, by semantically comparing it to the text being summarized, before presenting it to the user. This approach is embodied as the text summarizer HolSum, which is presented in this chapter and evaluated in Paper 5. A second aspect is the collection of large-scale corpora for languages where few or none such exist. This type of corpora is on the one hand needed for building the language model used by HolSum when comparing summaries on semantic grounds, on the other hand a large enough set of (written) language use is needed to guarantee the randomly selected subcorpus used for evaluation to be representative. This topic briefly touched upon in Chapter 4, and detailed in Paper 1. The third aspect is, of course, the evaluation of the proposed summarization method on a new language. This aspect is investigated in Chapter 4. Evaluations of HolSum have been run on English as well as on Swedish, using both well established data and evaluation schemes (English) as well as with corpora gathered “in the wild” (Swedish). During the development of the latter corpora, which is discussed in Paper 4, evaluations of a traditional sentence ranking text summarizer, SweSum, have also been run. These can be found in Paper 2 and 3. This thesis thus contributes a novel approach to highly portable automatic text summarization, coupled with methods for building the needed corpora, both for training and evaluation on the new language. / Idag, med ett överflöd av digitalt lagrad information även för många mindre språk, är det nära nog omöjligt att manuellt sålla och välja ut vilken information man ska ta till sig. Denna information måste istället filteras och extraheras för att man inte ska drunkna i den. En teknik för detta är automatisk textsammanfattning, där en dator sammanfattar en längre text till en kortare icke-redundant form. Vid sidan av de stora världsspråken finns det många små språk för vilka det saknas stora datamängder ämnade för språkteknologisk forskning. För dessa saknas det också ofta resurser för att bygga upp sådana datamängder då detta är tidskrävande och ofta dessutom kräver en ansenlig mängd manuellt arbete. Likväl behövs automatisk textsammanfattning för dessa språk för att tämja denna konstant ökande mängd elektronsikt producerad text. Denna avhandling sätter således fokus på automatisk sammanfattning av text med så liten mänsklig insats som möjligt. De använda resurserna bör i så hög grad som möjligt redan existera, inte behöva vara skapade för automatisk textsammanfattning och helst även ha kommit till som en naturlig del av en litterär process. Vidare, sammanfattningssystemet bör utan större ansträngning kunna sättas samman med hjälp av ett mindre antal mycket grundläggande språkteknologiska verktyg, vilka inte heller de är specifikt ämnade för textsammanfattning. Textsammanfattaren bör således vara nära nog språkoberoende för att det med enkelhet kunna att flyttas mellan ett språk och ett annat. Den forskning som läggs fram i denna avhandling berör i huvudsak tre datorsystem, ett för nära nog språkoberoende sammanfattning – HolSum; ett för insamlande av stora textmängder – KTH News Corpus; och ett för utvärdering av sammanfattning – KTH eXtract Corpus. Dessa tre system representerar tre olika aspekter av att föra över den framlagda sammanfattningsmetoden till ett nytt språk. En aspekt är den faktiska sammanfattningsmetoden och hur den påverkas av mänskliga språks högst oregelbundna natur och de skillnader som uppvisas mellan olika språkgrupper. Denna aspekt diskuteras i detalj i kapitel tre. I detta kapitel presenteras också begreppet “holistisk sammanfattning”, en ansats tillsjälvutvärderande sammanfattning vilken gör en innehållslig bedömning av sammanfattningen som en helhet innan den presenteras för användaren. Denna ansats förkroppsligas i textsammanfattaren HolSum, som presenteras i detta kapitel samt utvärderas i artikel fem. En andra aspekt är insamlandet av stora textmängder för språk där sådana saknas. Denna typ av datamängder behövs dels för att bygga den språkmodell som HolSum använder sig av när den gör innehållsliga jämförelser sammanfattningar emellan, dels behövs dessa för att ha en tillräckligt stor mängd text att kunna slumpmässigt extrahera en representativ delmängd lämpad för utvärdering ur. Denna aspekt berörs kortfattat i kapitel fyra och i mer önskvärd detalj i artikel ett. Den tredje aspekten är, naturligtvis, utvärdering av den framlagda sammanfattningsmetoden på ett nytt språk. Denna aspekt ges en översikt i kapitel 4. Utvärderingar av HolSum har utförts både med väl etablerade datamängder och utvärderingsmetoder (för engelska) och med data- och utvärderingsmängder insamlade specifikt för detta ändamål (för svenska). Under sammanställningen av denna senare svenska datamängd, vilken beskrivs i artikel fyra, så utfördes även utvärderingar av en traditionell meningsextraherande textsammanfattare, SweSum. Dessa återfinns beskrivna i artikel två och tre. Denna avhandling bidrar således med ett nydanande angreppssätt för nära nog språkoberoende textsammanfattning, uppbackad av metoder för sammansättning av erforderliga datamängder för såväl modellering av som utvärdering på ett nytt språk. / QC 20100712

No description available.

Leaching is one of the major processes occurring when organic litter is decomposed and is often completed within a few days when litter enters aquatic environments. It is important that leaching is addressed when studying microbial and invertebrate decomposition rates in order to avoid overestimations. The traditional litter bag method that has been used to measure decomposition rates in both terrestrial and aquatic environments has in recent years been challenged by the new and widely adopted tea bag index (TBI). Both methods, however, fail to bring a standardized methodology for separating and recognizing weight losses of litter due to leaching and biotic decomposition. Through a field experiment in two streams with different water discharge, this study has focused on exploring the leaching phase and post-leaching phase of the tea products used in TBI. The results unveiled that 20% of rooibos tea’s and 44% of green tea’s initial weight was lost to leaching within three days (72 hours) of the experiment. After the 72nd hour, both teas remained in a stabilized phase until the end of the experiment (120 hours). Water discharge had no significant effect on neither of the tea-weights during or after the leaching phase. This study recommends that weight loss through the leaching phase are taken into account in future studies and advocate the development of an updated TBI protocol where leaching losses are recognized. If not, overestimations of active decomposition rates will be made and may result in compromised conclusions. / Urlakning är en av de viktigaste processerna som uppstår vid nedbrytning av organiskt material och är ofta slutfört inom några dagar när materialet befinner sig i vattenmiljöer. Det är viktigt att den urlakade massan beräknas när studier fokuserade på nedbrytningshastigeter av mikrober och evertebrater genomförs, detta för att undvika en överestimering av den biotiska aktiviteten. Den traditionella metoden, där torkade växtdelar i påsar, har använts för att mäta nedbrytningshastigheter i både mark- och vattenmiljöer har under de senaste åren utmanats av det nya och allmänt accepterade tepåse-indexet (TBI). Båda metoderna misslyckas dock med att implementera en standardiserad metodik för att separera och uppskatta materialets viktförluster genom urlakning och biotisk nedbrytning. Genom ett fältexperiment i två vattendrag med olika vattenföring har denna studie fokuserat på att undersöka urlakningsfasen och den stabiliserade perioden efter urlakningen hos teprodukterna som används i TBI. Resultaten avslöjade att 20% av rooibos tes och 44% av grönt tes ursprungliga vikt förloras genom urlakning inom tre dagar (72 timmar) av experimentet. Efter den 72:a timmen förblev båda teerna i en stabiliserad fas till slutet av experimentet (120 timmar). Vattenflöde hade ingen signifikant effekt på någon av tetypernas vikter, varken för perioden under eller efter urlakningsfasen. Denna studie rekommenderar att viktminskning genom urlakningsfasen beaktas i framtida studier och förespråkar utvecklingen av ett uppdaterat TBI-protokoll för att inkludera urlakningens påverkan. Om inte, kommer överskattningar av aktiva nedbrytningshastigheter att göras och kan leda till äventyrade slutsatser.