Semantics for virtual humans

Alejandra Garcia Rojas Martinez
EPFL, 2009
EPFL thesis

Abstract

Population of Virtual Worlds with Virtual Humans is increasing rapidly by people who want to create a virtual life parallel to the real one (i.e. Second Life). The evolution of technology is smoothly providing the necessary elements to increase realism within these virtual worlds by creating believable Virtual Humans. However, creating the amount of resources needed to succeed this believability is a difficult task, mainly because of the complexity of the creation process of Virtual Humans. Even though there are many existing available resources, their reusability is difficult because there is not enough information provided to evaluate if a model contains the desired characteristics to be reused. Additionally, the knowledge involved in the creation of Virtual Humans is not well known, nor well disseminated. There are several different creation techniques, different software components, and several processes to carry out before having a Virtual Human capable of populating a virtual environment. The creation of Virtual Humans involves: a geometrical representation with an internal control structure, the motion synthesis with different animation techniques, higher level controllers and descriptors to simulate human-like behavior such individuality, cognition, interaction capabilities, etc. All these processes require the expertise from different fields of knowledge such as mathematics, artificial intelligence, computer graphics, design, etc. Furthermore, there is neither common framework nor common understanding of how elements involved in the creation, development, and interaction of Virtual Humans features are done. Therefore, there is a need for describing (1) existing resources, (2) Virtual Human's composition and features, (3) a creation pipeline and (4) the different levels/fields of knowledge comprehended. This thesis presents an explicit representation of the Virtual Humans and their features to provide a conceptual framework that will interest to all people involved in the creation and development of these characters. This dissertation focuses in a semantic description of Virtual Humans. The creation of a semantic description involves gathering related knowledge, agreement among experts in the definition of concepts, validation of the ontology design, etc. In this dissertation all these procedures are presented, and an Ontology for Virtual Humans is described in detail together with the validations that conducted to the resulted ontology. The goal of creating such ontology is to promote reusability of existing resources; to create a shared knowledge of the creation and composition of Virtual Humans; and to support new research of the fields involved in the development of believable Virtual Humans and virtual environments. Finally, this thesis presents several developments that aim to demonstrate the ontology usability and reusability. These developments serve particularly to support the research on specialized knowledge of Virtual Humans, the population of virtual environments, and improve the believability of these characters.

Official source

https://infoscience.epfl.ch/record/130368?ln=en

About this result

This page is automatically generated and may contain information that is not correct, complete, up-to-date, or relevant to your search query. The same applies to every other page on this website. Please make sure to verify the information with EPFL's official sources.

Related concepts

Related publications

Alejandra Garcia Rojas Martinez
EPFL, 2009
EPFL thesis

Abstract

Official source

https://infoscience.epfl.ch/record/130368?ln=en

About this result

Related concepts (14)

Related concepts

Related publications (10)

Related publications

Prior knowledge in Kernel methods

Alexei Pozdnoukhov

Machine Learning is a modern and actively developing field of computer science, devoted to extracting and estimating dependencies from empirical data. It combines such fields as statistics, optimization theory and artificial intelligence. In practical tasks, the general aim of Machine Learning is to construct algorithms able to generalize and predict in previously unseen situations based on some set of examples. Given some finite information, Machine Learning provides ways to exract knowledge, describe, explain and predict from data. Kernel Methods are one of the most successful branches of Machine Learning. They allow applying linear algorithms with well-founded properties such as generalization ability, to non-linear real-life problems. Support Vector Machine is a well-known example of a kernel method, which has found a wide range of applications in data analysis nowadays. In many practical applications, some additional prior knowledge is often available. This can be the knowledge about the data domain, invariant transformations, inner geometrical structures in data, some properties of the underlying process, etc. If used smartly, this information can provide significant improvement to any data processing algorithm. Thus, it is important to develop methods for incorporating prior knowledge into data-dependent models. The main objective of this thesis is to investigate approaches towards learning with kernel methods using prior knowledge. Invariant learning with kernel methods is considered in more details. In the first part of the thesis, kernels are developed which incorporate prior knowledge on invariant transformations. They apply when the desired transformation produce an object around every example, assuming that all points in the given object share the same class. Different types of objects, including hard geometrical objects and distributions are considered. These kernels were then applied for images classification with Support Vector Machines. Next, algorithms which specifically include prior knowledge are considered. An algorithm which linearly classifies distributions by their domain was developed. It is constructed such that it allows to apply kernels to solve non-linear tasks. Thus, it combines the discriminative power of support vector machines and the well-developed framework of generative models. It can be applied to a number of real-life tasks which include data represented as distributions. In the last part of the thesis, the use of unlabelled data as a source of prior knowledge is considered. The technique of modelling the unlabelled data with a graph is taken as a baseline from semi-supervised manifold learning. For classification problems, we use this apporach for building graph models of invariant manifolds. For regression problems, we use unlabelled data to take into account the inner geometry of the input space. To conclude, in this thesis we developed a number of approaches for incorporating some prior knowledge into kernel methods. We proposed invariant kernels for existing algorithms, developed new algorithms and adapted a technique taken from semi-supervised learning for invariant learning. In all these cases, links with related state-of-the-art approaches were investigated. Several illustrative experiments were carried out on real data on optical character recognition, face image classification, brain-computer interfaces, and a number of benchmark and synthetic datasets.

EPFL2006

École Polytechnique Fédérale de Lausanne2006

In combination with the rapid technical improvements of computers, building large virtual scenes has become a popular field in computer graphics for education, films or games. Often, virtual humans populate scenes such as real reconstructed city, disappeared building or virtual town, and they are expected to provide a real life feeling. The specific aim of this thesis is to perform action planning in order to populate these scenes with virtual humans. To achieve this goal, we use a database dedicated to urban life simulation called an Informed Environment. This report presents a design methodology and associated tools for animation purposes in large scale scenes. We propose a space decomposition of the scene into a set of entities allowing automatic path computing regardless of scene size. In order to achieve human behavioural simulation, we decide to focus on the definition of scene decomposition resulting in the creation of a database. The first stage of this work consisted of building a database simulating "urban knowledge" (knowledge about location and associated functionalities) and virtual perceptions (contact with the ground, object perception etc.). These virtual perceptions, necessary for autonomy simulation, but highly costly in time processor consuming, are simulated through database access corresponding to our Informed Environment. We decided to define an ideal 3D model as a test-bed for concepts and tools, integrating both geometrical and semantical data. We also define rules for the design, storage and display options. The "urban knowledge" attached to a simulation is the base of a generic hierarchical model of space decomposition. This model of space decomposition of the 3D scene is related to a specific type of simulation and directly applied to the modelling of the tridimensional scene. Once the database is created, some tools access the database content for path/action computation, knowledge concerning a place, display optimization, places and objects. As 3D data visibility and verification with associated semantic information are difficult to observe, we have defined some tools for database visualisation, verification and modification (the "applicity" software). These tools have been developed in order to offer assistance to the user during manipulation of three-dimensional data. This system is applied to real reconstructed scenes in the framework of the European project CROSSES, for panic simulation during accident. Our field of research has also been extended to the domain of disability help in order to compute path for people with reduced mobility. Our goal was to define some tools and a few constraint methodology that automatically generate all information needed for simulating autonomous behaviour in an urban environment.

EPFL2001

Prior knowledge in Kernel methods

Alexei Pozdnoukhov

EPFL2006

EPFL2001

École Polytechnique Fédérale de Lausanne2006