305 research outputs found
Continuous Interaction with a Virtual Human
No full textReidsma D, Truong K, van Welbergen H, et al. Continuous Interaction with a Virtual Human. In: Salah AA, Gevers T, eds. Proceedings eNTERFACE'10: 6th International Summer Workshop on Multimodal Interfaces (eNTERFACE'10). Amsterdam: University of Amsterdam; 2010: 24-39
A proposal for image indexing: keypics, plastic graphical metadata
No full textWe propose a graphical indexing of images to be exposed on the Web. This should be accomplished by “keypics”,
i.e. auxiliary, simplified pictures referring to the geometrical and/or the semantic content of the indexed image.
Keypics should not be rigidly standardized; they should be left free to evolve, to express nuances and to
stress details. A mathematical tool for dealing with such freedom already exists: Size Functions.
We support the idea of keypics with some experiments on a 498 images dataset
A proposal for image indexing: keypics, plastic graphical metadata
No full textWe propose a graphical indexing of images to be exposed on the Web. This should be accomplished by “keypics”,
i.e. auxiliary, simplified pictures referring to the geometrical and/or the semantic content of the indexed image.
Keypics should not be rigidly standardized; they should be left free to evolve, to express nuances and to
stress details. A mathematical tool for dealing with such freedom already exists: Size Functions.
We support the idea of keypics with some experiments on a 498 images dataset
Methods and technologies for gait analysis
No full textGait analysis is a highly active research area with a wide range of applications in clinical settings, surveillance and human-computer interaction. The focus of this chapter is the clinical aspect of gait analysis, in which accuracy and precision are essential. Subsequently, the chapter focuses on various techniques of measuring gait and introduces taxonomy for their analysis. From this perspective, motion measurements using motion capture and inertial sensors are presented. Motion capture techniques are analyzed under sections of marker-based and markerless techniques and their common applications are exemplified. Additionally, accelerometers, gyroscopes, magnetometers and their applications are presented in the inertial measurements section. Finally, force measurements and measurement of electrical activity of muscles are explained briefly
Description and evaluation of techniques for transfer learning across sub-categories
No full textThis report presents contributions in sub-categorisation. In the first part we propose a simple approach to build a meta-learner on individual classifier that is supposed to implicitly learn class inter-dependencies. We further study how the performance is modified if we add automatically learned meta-classes (group of similar classes). In the second part, we address the problem using a structured learning approach. The main idea is to build a graphical model which efficiently performs labelling at multiple levels simultaneously
Facial features matching using a virtual structuring element
No full textFace analysis in a real-world environment is a complex task as it should deal with challenging problems such as pose variations, illumination changes and complex backgrounds. The use of active appearance models for facial features detection is often successful in restricted environments, but the performance decreases when applied in unconstrained environments. Therefore, in this paper, we introduce a novel method that integrates the knowledge of a face detector inside the shape and the appearance models by using what we call a 'virtual structuring element' (VSE). In this way the possible settings of the active appearance models are constrained in an appearance-driven manner. The use of a virtual structuring element in an active appearance model provides increased performance in both accuracy and robustness over standard active appearance models applied to different environments
Continuous Analysis of Affect from Voice and Face
No full textHuman affective behavior is multimodal, continuous and complex. Despite major advances within the affective computing research field, modeling, analyzing, interpreting and responding to human affective behavior still remains a challenge for automated systems as affect and emotions are complex constructs, with fuzzy boundaries and with substantial individual differences in expression and experience [7]. Therefore, affective and behavioral computing researchers have recently invested increased effort in exploring how to best model, analyze and interpret the subtlety, complexity and continuity (represented along a continuum e.g., from −1 to +1) of affective behavior in terms of latent dimensions (e.g., arousal, power and valence) and appraisals, rather than in terms of a small number of discrete emotion categories (e.g., happiness and sadness). This chapter aims to (i) give a brief overview of the existing efforts and the major accomplishments in modeling and analysis of emotional expressions in dimensional and continuous space while focusing on open issues and new challenges in the field, and (ii) introduce a representative approach for multimodal continuous analysis of affect from voice and face, and provide experimental results using the audiovisual Sensitive Artificial Listener (SAL) Database of natural interactions. The chapter concludes by posing a number of questions that highlight the significant issues in the field, and by extracting potential answers to these questions from the relevant literature. The chapter is organized as follows. Section 10.2 describes theories of emotion, Sect. 10.3 provides details on the affect dimensions employed in the literature as well as how emotions are perceived from visual, audio and physiological modalities. Section 10.4 summarizes how current technology has been developed, in terms of data acquisition and annotation, and automatic analysis of affect in continuous space by bringing forth a number of issues that need to be taken into account when applying a dimensional approach to emotion recognition, namely, determining the duration of emotions for automatic analysis, modeling the intensity of emotions, determining the baseline, dealing with high inter-subject expression variation, defining optimal strategies for fusion of multiple cues and modalities, and identifying appropriate machine learning techniques and evaluation measures. Section 10.5 presents our representative system that fuses vocal and facial expression cues for dimensional and continuous prediction of emotions in valence and arousal space by employing the bidirectional Long Short-Term Memory neural networks (BLSTM-NN), and introduces an output-associative fusion framework that incorporates correlations between the emotion dimensions to further improve continuous affect prediction. Section 10.6 concludes the chapter
Towards Deep Image Understanding : from pixels to semantics
No full textEntendre el contingut de les imatges és un dels grans reptes de la visió per computador. Arribar a ser capaços de reconèixer quins objectes apareixen en les imatges, quina acció hi realitzen, i finalment, entendre el per què esta succeïnt, és l'objectiu del topic de Image Understanding. El fet d'entendre què succeeix en un instant de temps, ja sigui capturat en una fotografia, en un vídeo o simplement la imatge retinguda en la retina de l'ull (humà o un robòtic) és un pas fonamental per tal de formar-n'hi part. Per exemple, per un robot o un cotxe intel·ligent, es imprescindible de reconèixer el que succeeix en el seu entorn per tal de poder-hi navegar i interactuar de forma segura. O bé, es pot interactuar amb el contingut d'una imatge i extreure'n conceptes textuals per desprès ser utilitzats en els buscadors d'Internet actuals. En aquesta tesis es pretén descobrir què apareix en una imatge, i com extreure'n informació semàntica de més alt nivell. En altres paraules, l'objectiu és el de categoritzar i localitzar els objectes dins d'una imatge. Abans de res, per tal d'aprofundir en el coneixement sobre la formació d'imatges, proposem un mètode que aprèn a reconèixer alguna de les propietats físiques que han creat la imatge. Combinant informació fotomètrica i geomètrica, aprenem a dir si un gradient ha estat format pel material de l'objecte dins l'escena o bé si ha estat causat per alteracions a l'escena com ombres o reflexos. Endinsant-nos en l'àmbit del reconeixement semàntic dels objectes, ens centrem en dues aproximacions per a descriure els objectes. En la primera volem reconèixer quina categoria d'objecte s'amaga darrera de cada píxel, el que s'anomena segmentació semàntica. La segona aproximació s'inclou dins el tòpic de detecció d'objectes, en el que no són tan important els píxels, sinó l'objecte sencer i es es representa a través d'un requadre envoltant l'objecte. La segmentació semàntica és un problema en el que la ambigüitat dels píxels s'ha de resoldre a través d'afegir característiques contextuals. Nosaltres proposem que el context a varis nivells d'escala s'ha de tractar de forma diferent. A baix nivell ens podem aprendre si l'aparen\c{c}a d'un píxel podria representar l'objecte o no, però per estar-ne més segurs es requereix de més informació. En els metodes que proposem, incloim la informació de entitat i la coherencia amb la resta de l'escena, introduint la co-ocurrència semàntica. Pel que fa a la detecció d'objectes, es proposen dos nous algoritmes. El primer, es basa en millorar la representació d'objectes a nivell local, introduint el concepte de factorització d'aparences. D'aquesta manera, un objecte esta representat per diferents parts, i cada una de les parts podria ser representada per més d'una aparen\c{c}a. Finalment, l'últim mètode proposat adre\c{c}a el problema computacional de reconèixer i localitzar milers de categories d'objectes en una imatge. El principi bàsic és el de crear representacions d'objectes que siguin útils per qualsevol tipus d'objecte, i així reaprofitar la computació de la representació.Entender el contenido de las imágenes es uno de los grandes retos de la visión por computador. Llegar a reconocer cuales son los objetos que aparecen en las imágenes, qué acciones están realizando, y finalmente, entender el porqué sucede, es el objetivo del tópico de "Image Understanding". El hecho de entender que está sucediendo en un tiempo determinado, ya sea mediante la toma de una fotografía, en un video, o simplemente la imagen reflejada en la retina del ojo (humano o robótico) es una paso fundamental para llegar a formar parte de ese instante. Por ejemplo, para un robot o coche inteligente, es imprescindible reconocer que sucede al su alrededor para poder navegar y interactuar con el entorno de forma segura. Otro ejemplo se puede encontrar en el hecho de interactuar con el contenido de las imágenes, de modo que se puedan extraer conceptos textuales de esta, para luego ser utilizados en los buscadores de Internet actuales. En esta tesis se pretende descubrir que aparece en una imagen, y como se puede extraer información semántica de mas alto nivel. En otras palabras, el objetivo es el de categorizar y localizar los objetos dentro de una imagen. Antes de nada, para profundizar en el conocimiento sobre la formación de las imágenes, proponemos un método que aprende a reconocer las propiedades físicas que han creado la imagen. Combinando información fotométrica y geométrica, podemos aprender a decir si un gradiente ha sido creado por variaciones en el materiales de los objetos o bien, si es causado por alteraciones en la escena como sombras o reflejos. Entrando en el ámbito del reconocimiento semántico de los objetos, nos centramos en dos aproximaciones para describir los objetos. En la primera, queremos reconocer qué categoría de objeto se esconde detrás de los pixeles, lo que denominamos segmentación semántica de imágenes. La segunda aproximación se incluye en el tópico de detección de objetos}, en el que no es tan importante el resultado en los pixeles, sino dónde se encuentra un objeto entero. Se representa a través de un recuadro que envuelve el objeto. La segmentación semántica es un problema en el que la ambigüedad de los pixeles se debe resolver a través de añadir características contextuales. Nosotros proponemos que el contexto a varios niveles de escala se debe tratar de forma distinta. A bajo nivel, podemos aprender si la apariencia de un pixel podría parecerse a la del objeto o no, pero para estar seguros se requiere mas información. En los métodos que proponemos, añadimos información del objeto como entidad y la coherencia con el resto de la escena, introduciendo el concepto de co-ocurréncia semántica. En cuanto a la detección de objetos, se proponen dos nuevos algoritmos. El primero, se basa en mejorar la representación de los objetos a nivel local, con el concepto de factorización de apariencias. De este modo, un objeto se representa con varias partes, y cada una de las partes puede ser representada por más de una apariencia. Finalmente, el último método propuesto aborda el problema computacional de reconocer y localizar miles de categorías de objetos en una imagen. El principio básico es el de crear representaciones que objetos que sean útiles para cualquier tipo de objeto, y así reaprovechar la computación de la representación.Understand the content of the images is one of the great challenges of computer vision. Being able to recognize which are the objects in the images, what actions are doing, and finally understand why it happens, is the purpose of Image Understanding. The fact of understanding what is happening in a given time, either by taking a picture, video, or simply the image on the retina of the eye (human or robot) is a fundamental step to become part of that instant. For example, for a robot or smart car is essential to recognize what is succeeding to navigate around and interact with the environment safely. Another example can be found by interacting with the image content, so that their textual concepts can be used in modern Web searchers. This thesis seeks to discover what appears in a picture, and how to extract semantic information of higher level. In other words, the objective is to categorize and locate objects within an image. First of all, to deepen the knowledge on the formation of images, we propose a method that learns to recognize the physical properties that have created the image. By combining photometric and geometric information, we can learn to say whether a gradient is created by variations in the materials or objects, or it is caused by alterations in the scene as shadows or reflections. Entering the field of semantic recognition of objects, we focus on two approaches to describe the objects. First, we recognize which object category is hidden behind the pixels, which we call semantic segmentation. The second approach is included in the topic of object detection, which is not as important outcome in pixels, but where there is a whole object. Is represented by a frame which surrounds the object. Semantic segmentation is a problem in which the ambiguity of the pixels must be resolved by adding contextual features. We propose that the context at various scale levels should be treated differently. At low level, we learn whether the appearance of a pixel resembles the object or not, but to become confident, more information is required. We add information about the object as an entity and we enforce consistency with the rest of the scene, introducing the concept of semantic co-occurrence. As for object detection, we propose two new algorithms. The first is based on improving the representation of objects locally, with the concept of factorize appearances. Thus, an object is represented by several parts, and each of the parts can be represented by more than one appearance. Finally, the last proposed method addresses the computational problem of identifying and locating thousands of categories of objects in an image. The basic principle is to create representations of objects that are useful for any type of object, and thus reuse the computation of the performance
‘Germanen sprechen französisch, Romanen fühlen flandrisch’? German translations of Marie Gevers under National-Socialism
No full textThis article will study the issue of the German translations of Belgian/Flemish author Marie Gevers, which were published under National-Socialism. These translations were part of a long-term program meant to promote positive images of Flanders in German-speaking countries. Paradoxically the Nazis tolerated the presence of a French-language writer, while they were encouraging the social emancipation of the Flemish Movement to liberate itself from the socio-cultural dominance of French in Belgium. Close-readings of Gevers’s texts will shed light on this specific contradiction and the way the regional literature embodied by her novels may have suited the propagandistic model of the ‘völkisch’ literature promoted by the Nazis. Another aspect of my interpretations concerns the images of Flanders in Gevers’s German translations and how they favor a kind of aporia through some topoi and hetero-representations of Belgium, which had been common in Germany and in Europe since the nineteenth century
Geven door bedrijven
No full text• Bedrijven geven naar schatting E1,9 miljard in 2018• Sponsoring en giften door bedrijven per doel• Robuuste meerjarige trends• Grote verschillen tussen branches in gegeven bedragen• Gulle gevers versus brede steun• Veel bedrijven geven helemaal niet• Geven in geld, natura en mankracht• Motieven voor sponsoring en giften• Beleid van bedrijven voor sponsoring en giften• Maatschappelijk Verantwoord Ondernemen (MVO)• Motieven om maatschappelijk verantwoord te onderneme
- …
