|
|
Model V1 s realistickou distribucí funkčních typů neuronů v rámci kortikálních vrstev
Moudřík, Josef ; Sýkora, Ondřej (referee) ; Brom, Cyril (advisor)
There have been identified two functionally different neuron classes in the primary visual cortex (V1), so-called simple and complex cells. These cells differ in reactions to various stimuli and their development has been successfully simulated in one computational model of V1. This model, however, simulates both classes in separate layers corresponding to layers 4C and 2/3 in V1. On the contrary, experiments have shown that both categories are - in different proportions - present in both layers. In this thesis, a computational model with a realistic distribution of complex and simple cells is presented. To increase its authenticity, I incorporate long-range excitatory and short-range inhibitory lateral cortical connections as found in animals, overcoming one drawback of previous models that used long-range inhibition. To assess my model, two measures of orientation selectivity - circular variance and orientation bandwidth - were computed for each simulated neuron. Using this measures, I compared my model with data from macaque monkey. In line with biological findings, my model develops a wide diversity of orientation selectivity. Moreover, it develops maps of orientation preference and realistic receptive fields.
|
|
|
Activity of V1 neurons after presentation of novel class of stimuli: comparision with experimental data
Machálek, Jiří ; Lukavský, Jiří (referee) ; Brom, Cyril (advisor)
Neurons in visual cortex areas V1 and V2 are selective to the shape of stimulus. Almost all present theories and models assume that variability to the shape of stimuli to which the neurons are selective is very low - preferred shapes are typically only bars and edges. However recent study (Hegdé and Van Essen) showed that is not true { single neurons prefer wide range of shapes. This work replicates the study in current computational models of cortical areas V1 and V2 (for example LISSOM) and compares results with experimental data. It shows that examined models develope neurons preferring these complex shapes rarely or just for a part of stimuli set.
|
|
|
Gestures and facial expression of virtual humans in 3D worlds
Píbil, Radek ; Horáček, Jan (referee) ; Brom, Cyril (advisor)
Goal of this work is to explain design and implementation of an extension for Pogamut that enables user's virtual beings (agents) to control its avatar's (agent's body) animation on a more complex level than unmodied Unreal Tournament 2004 (UT2004) and Pogamut allow. Pogamut is a platform for research and creation of agents and UT2004 is an environment used for their simulation. Gestures and facial expressions, animation in general, are key elements of believability of agents in virtual world. Thus animation is one of the directions in which Pogamut is evolving. This thesis describes communication of this extension between Pogamut and UT2004, handling of animation that avatars are intended to execute and compares dierent approaches to realization of each part of this extension.
|
|
|
Debugging tools in Pogamut 2
Havlíček, Jan ; Bureš, Tomáš (referee) ; Brom, Cyril (advisor)
Virtual agent is a program which is trying to achieve its goals in a virtual environment based on an available information from the environment and its decision mechanism. Project Pogamut is platform for creation of such agents in an environment of game Unreal Tournament. While Pogamut makes it easy to create logic for the agents, its debugging capabilities are lacking. Goal of this thesis is to identify disadvantages of current debugging tools used in Pogamut and to describe and create a new debugging tool that would replace former debugging tools and that would be more easy to use.
|
|
|
Egocentric module for episodic memory
Vyhnánek, Jan ; Sýkora, Ondřej (referee) ; Brom, Cyril (advisor)
The present research in cognitive psychology concerns with the issue of how human beings remember positions of objects in space - and what mental representation of space they create. Scientists infer from experiments that human beings create several spatial representations which differ in accuracy and rate of forgetting and which are specified with respect to different spatial frames of reference. However, we do not know how exactly these representions work. As far as we know there is no computational model concerning with this issue, which would be useful for understanding of these representations. The goal of this work is to extend the model of episodic-like memory developed at MFF UK with a mechanism of creating spatial representations. We will examine how the representations are created and which representations are the most important for orientation of a human-like agent. The model was tested on simulations of a psychological experiment and it partially predicted its results. The goal of the model is not to give a full answer to this problem. It is supposed to be an attempt for starting research in the area which has not been researched this way yet. The results of this work may contribute to better understanding of how human beings represent positions of objects and it may be useful for designing...
|
|
|
Taktická hra pre viacej hráčov
Kis, Ján ; Sýkora, Ondřej (referee) ; Brom, Cyril (advisor)
The tactics single-player games like UFO from Altar Interactives are quite popular in these days. The most notable feature of UFO is that the actions of game units can be planned in advance by the player while the game is paused. During the recent boom of multiplayer games a question arose of what should a multiplayer to a game such as UFO look like. This project addresses the question and discusses several possibilities of the multiplayer design to such a game. The main contribution of the project is the actual game prototype which tests the viability and entertainment of the multiplayer design introduced by the thesis.
|
|
|
HTN planning for project IVE
Holeček, Ondřej ; Brom, Cyril (advisor) ; Plch, Tomáš (referee)
IVE is an ambitious project whose main effort was to create a highly scalable architecture to allow discrete simulation of large-scale virtual world full of virtual beings and their virtual lives. Part of the project is an implementation of reactive virtual beings fully driven by their environment. In this work I try to extend this reactive intelligence by ability to plan in advance actions and their specific execution order to successfully accomplish its tasks. I am trying to achieve this by attaching external HTN planning system, implementing automatic world representation translation and mutual interaction between IVE framework and planning subsystem. Along with planning I add not only ability to plan ones individual tasks, but also more complex planning of cooperation of multiple beings on common goals, which may be somewhat difficult using only reactive control.
|
|
|
Memory model for a human-like agent
Pešková, Klára ; Neruda, Roman (referee) ; Brom, Cyril (advisor)
In this thesis we introduce a model of episodic memory for a human-like agent, who is designed according to Gibson's theory of affordances. Episodic memory represents personal history of an agent. It stores memories related to particular places and moments, their storage and retrieval depends on subjective feelings and current goals. In this thesis we discuss the believability of the agent as well as some technical issues concerning memory size and memory access time. As a part of our work, we implemented a human-like agent with hierarchical reactive planning, including some features of Belief-Desire-Intention. The agent remembers only important things, i.e. things which passed through the attention filter. Thanks to this filter, memory size can be significantly reduced. At the end of the thesis we introduce results of some tests, which were performed in our prototype implementation.
|
|
|
Adaptation of spatial navigation tests to virtual reality.
Šupalová, Ivana ; Holan, Tomáš (referee) ; Brom, Cyril (advisor)
At the Department of Neurophysiology of Memory in the Academy of Sciences in Czech Republic are recently performed tests of spatial navigation of people in experimental real enviroment called Blue Velvet Arena. In introdution of this thesis is described importancy of these tests for medical purposes and the recent solution. The main aim is to adapt this real enviroment to virtual reality, allow it's configuration and enable to collect data retieved during experiment's execution. Resulting system is using cooperation of Unreal engine, application in Java language and script engine. The work is also enabling to use Eyelink II device for measuring eye movements.
|
|
|
Mechanismus výběru akcí pro animata
Plch, Tomáš ; Surynek, Pavel (referee) ; Brom, Cyril (advisor)
In this thesis we study the method of reactive planning, which is very popular for modeling behavior of virtual beings. Advantage of this approach is in the simplicity of design, its speed and believable behavior of agents in dynamic and complex environments. Thanks to the simplicity of the concept, it is rather difficult to model complex forms of behavior in a way, that the resulting behavior is believable and perceived intelligent. This thesis is focused on the expanding the concept of reactive planning in such a fashion, to compensate for observed limitations, preserving the basic concepts. The thesis contains proposals for various improvements to the basic structure of reactive plans - the reactive plan. In the presented work, we analyze practical problem of action selection used in reactive planning. Part of the thesis is a simple prototype, which is used to present some of the tackled limitations of reactive planning.
|
guest ::
