Lo visual como construcción. Desarrollo de un modelo pictórico de representación de la imagen

903 páginas, bibliografía en páginas 854-895, glosario en páginas 896-903

Aportación a las representaciones de úrsidos en el arte mobiliar magdaleniense

[ES] Durante la primera mitad del siglo XX se realizaron amplias excavaciones arqueológicas en la cueva de Isturitz que aportaron abundantes evidencias industriales y artísticas. En este artículo se trata de dar a la luz una obra de arte mueble que permanece inédita, en concreto una pareja de osos fabricados sobre una costilla, y que procede del nivel magdaleniense de la colección Saint-Périer.

Visual SLAM using straight lines

The present thesis is focuses on the problem of Simultaneous Localisation and Mapping (SLAM) using only visual data (VSLAM). This means to concurrently estimate the position of a moving camera and to create a consistent map of the environment. Since implementing a whole VSLAM system is out of the scope of a degree thesis, the main aim is to improve an existing visual SLAM system by complementing the commonly used point features with straight line primitives. This enables more accurate localization in environments with few feature points, like corridors. As a foundation for the project, ScaViSLAM by Strasdat et al. is used, which is a state-of-the-art real-time visual SLAM framework. Since it currently only supports Stereo and RGB-D systems, implementing a Monocular approach will be researched as well as an integration of it as a ROS package in order to deploy it on a mobile robot. For the experimental results, the Care-O-bot service robot developed by Fraunhofer IPA will be used.

Skeleton based visual pattern recognition. Applications to tabletop interaction

211 p. :il.

Deep learning review and its applications

Deep neural networks have recently gained popularity for improv- ing state-of-the-art machine learning algorithms in diverse areas such as speech recognition, computer vision and bioinformatics. Convolutional networks especially have shown prowess in visual recognition tasks such as object recognition and detection in which this work is focused on. Mod- ern award-winning architectures have systematically surpassed previous attempts at tackling computer vision problems and keep winning most current competitions. After a brief study of deep learning architectures and readily available frameworks and libraries, the LeNet handwriting digit recognition network study case is developed, and lastly a deep learn- ing network for playing simple videogames is reviewed.