Search

Este trabalho de Iniciacao Cientifica teve como objetivo principal o uso do classificador OPF (Optimum-Path Forest) para a caracterizacao de emocoes faciais a partir da base CH-Unicamp de videos. Uma vez adaptada para uma base de imagens e escolhidas caracteristicas, diferentes classificadores foram treinados, testados e comparados ao OPF.

This article presents results of the simulation of SIFT based algorithms in the context of the identification of tattoos. The algorithms studied are the SIFT - Scale Invariant Feature Transform, ASIFT - Affine SIFT, BOV - Bag of Visual Words and FV - Fisher Vector. The use of the OPF - Optimum-Path Forest and SVM - Support Vector Machine classifiers is exploited in conjunction with SIFT and ASIFT algorithms as well as BOV and FV. The present study uses the National Institute of Standards and Technology (NIST) Tatt-C dataset in a reduced and complete version. This work uses runtime and accuracy to compare the results of the simulations.

In content-based image retrieval (CBIR) using feedback-based learning, the user marks the relevance of returned images and the system learns how to return more relevant images in a next iteration. In this learning process, image comparison may be based on distinct distance spaces due to multiple visual content representations. This work improves the retrieval process by incorporating multiple distance spaces in a recent method based on optimum-path forest (OPF) classification. For a given training set with relevant and irrelevant images, an optimization algorithm finds the best distance function to compare images as a combination of their distances according to different representations. Two optimization techniques are evaluated: a multi-scale parameter search (MSPS), never used before for CBIR, and a genetic programming (GP) algorithm. The combined distance function is used to project an OPF classifier and to rank images classified as relevant for the next iteration. The ranking process takes into account relevant and irrelevant representatives, previously found by the OPF classifier. Experiments show the advantages in effectiveness of the proposed approach with both optimization techniques over the same approach with single distance space and over another state-of-the-art method based on multiple distance spaces.

This paper presents a method for crowd simulation based on a biologically motivated space colonization algorithm. This algorithm was originally introduced to model leaf venation patterns and the branching architecture of trees. It operates by simulating the competition for space between growing veins or branches. Adapted to crowd modeling, the space colonization algorithm focuses on the competition for space among moving agents. Several behaviors observed in real crowds, including collision avoidance, relationship of crowd density and speed of agents, and the formation of lanes in which people follow each other, are emergent properties of the algorithm. The proposed crowd modeling method is free-of-collision, simple to implement, robust, computationally efficient, and suited to the interactive control of simulated crowds.

This paper discusses methods for content-based image retrieval (CBIR) systems based on relevance feedback according to two active learning paradigms, named greedy and planned. In greedy methods, the system aims to return the most relevant images for a query at each iteration. In planned methods, the most informative images are returned during a few iterations and the most relevant ones are only presented afterward. In the past, we proposed a greedy approach based on optimum-path forest classification (OPF) and demonstrated its gain in effectiveness with respect to a planned method based on support-vector machines and another greedy approach based on multi-point query. In this work, we introduce a planned approach based on the OPF classifier and demonstrate its gain in effectiveness over all methods above using more image databases. In our tests, the most informative images are better obtained from images that are classified as relevant, which differs from the original definition. The results also indicate that both OPF-based methods require less user involvement (efficiency) to satisfy the user's expectation (effectiveness), and provide interactive response times.

This paper deals with the behavior of virtual environments from the collaboration point-of-view, in which actors (human or virtual beings) interact and collaborate by means of interdependent tasks. In this sense, actors may realize tasks that are dependent on tasks performed by other actors, while the interdependencies between tasks (through resource management and temporal relations) delineate the overall behavior of a virtual environment. Our main goal is to propose an approach for the coordination of those behaviors. Initially a generic study of possible interdependencies between collaborative tasks is presented, followed by the formal modeling (using Petri Nets) of coordination mechanisms for those dependencies. In order to implement such mechanisms, an architecture of reusable and pluggable coordination components is also introduced. These components are used in an implementation of a multi-user videogame. The presented approach is a concrete step to create virtual societies of actors that collaborate to reach common goals without the risk of getting involved in conflicting or repetitive tasks. r 2001 Elsevier Science Ltd. All rights reserved.

This paper introduces the use of Petri Nets as a modeling and analysis tool for animation environments. Firstly, the original formulation for Petri Nets is applied in two animation situations, one modeled as a state machine and another exploring interdependent transitions. Increasing the complexity level, some modeling extensions are discussed and more sophisticated animation examples are studied.

In this paper, a reference model aiming to contribute to the discussion and formalization of the issues related to the integration of Computer Graphics (CG), Image Processing (IP), and Computer Vision (CV) is proposed. This model takes into account that CG, IP, and CV are overlapping areas, and that CG and CV have inverse objectives. Regarding existing and forthcoming standards, the proposed model addresses the problems of data representation, data reduction, and communication. The model is based on three main issues: abstraction of data into layers, user dialog, and information interchange. It fulfills the requirements from various application areas including multimedia.

The use of remote sensing images as a source of information in agribusiness applications is very common. In those applications, it is fundamental to know how the space occupation is. However, identification and recognition of crop regions in remote sensing images are not trivial tasks yet. Although there are automatic methods proposed to that, users very often prefer to identify regions manually. That happens because these methods are usually developed to solve specific problems, or, when they are of general purpose, they do not yield satisfying results. This work presents a new interactive approach based on relevance feedback to recognize regions of remote sensing. Relevance feedback is a technique used in content-based image retrieval (CBIR) tasks. Its objective is to aggregate user preferences to the search process. The proposed solution combines the Optimum-Path Forest (OPF) classifier with composite descriptors obtained by a Genetic Programming (GP) framework. The new approach has presented good results with respect to the identification of pasture and coffee crops, overcoming the results obtained by a recently proposed method and the traditional Maximimun Likelihood algorithm.

This article presents an approach for generating steering behaviors of groups of characters based on the space colonization algorithm that has been used in the past for generating leaf venation patterns and tree structures. In this article, the underlying idea of the space colonization algorithm is adapted to control the motion of virtual characters, providing robust and realistic group behaviors by adjusting just a few parameters. The main contributions of this work are the robustness, flexibility, and simplicity of the proposed approach to control groups of characters in an interactive way, providing path planning and a series of group behaviors, such as group formation, alignment among others. We also introduce a possible extension of this model to provide collision avoidance among agents, mainly focused on crowd simulation. In addition, an interactive tool is provided to allow an easy manner for controlling the motion of virtual characters.

This paper describes a model for generating steering behaviors of groups of characters based on the biologically-motivated space colonization algorithm. This algorithm has been used in the past for generating leaf venation patterns and tree structures, simulating the competition for space between growing veins or branches. Adapted to character animation, this model is responsible for the motion control of characters providing robust and realistic group behaviors by adjusting just a few parameters. The main contributions are related with the robustness, flexibility and simplicity to control groups of characters.

This paper presents Relations Graph - GR a methodology to automate the generation of coordination mechanisms in computational environments. GR explores encapsulation and compacting capabilities of Colored Petri Nets to generate temporal coordination mechanisms, although the use of the GR methodology does not depend on the knowledge of PN formalism. GR supports alternative temporal behaviors and alternative activities changing the temporal relations among activities in processing time. An algorithm to identify and model coordination mechanisms linear to the number of activities and its application to an illustrative collaborative authoring environment will be presented.

We have developed a Web based application to accelerate the visualization of VRML scenes located in a remote server. This application enables the user to extract only the parts of a scene that are of actual interest. The extracted parts represent one or more sub trees of the hierarchical structure of the VRML scene, and only these parts will be rendered and visualized in the local computer. By reducing the complexity (size) of the remote scene, less data are transmitted from the remote server and the rendering process becomes faster in the local computer. The application is written in Java and is executed as an applet embedded in an HTML page.

The paper exploits the combination of VRML (Virtual Reality Modeling Language) and Java for the construction of highly interactive animations, whose behaviour is defined in real time by a user's actions. The animations are modeled in VRML, which allows the definition of a Java program to process and generate events that determine the behaviour of scene elements. An application for the generation of Java graphical interfaces was developed, aiming to establish the communication between the user and the VRML environment, sending parameters to the program that controls the animation.

This paper presents a fuzzy Petri net based approach suitable for the modeling of flexible coordination mechanisms to deal with temporal interdependencies between collaborative tasks. Such approach is based on an extension of the generalized fuzzy Petri net model, including the notion of time for the execution and synchronization of these tasks. A scenario of study is described, indicating the suitability of the proposal.

In this paper we explore the use of Petri nets as a tool to control the movements of articulated figures in computer animations. This approach permits us to describe the animation sequence by means of the treatment of events present in its execution. An advantage of this method is that the control may be abstracted in different levels, spanning from the definition of the relation among limbs for a single movement to behavioral directives. In addition, our treatment of events hides the mathematical model that describes the movement in fact, allowing the animators to choose the better technique for their applications. In this paper we use an inverse kinematics tool for this purpose. The use of Petri nets also allows previewing the behavior of the animation before starting any shot.

In this work the architecture of an animation machine is proposed. Its main goals are the definition of the animation in a more abstract level and the simulation of the characters' behaviors to detect undesired situations.

In this work we present a methodology to express both analytically and graphically the interdependencies among tasks realized in a collaborative environment. For each interdependency expression, a coordination mechanism is built, modeling the global behavior of the environment, i.e., the structure that ensures the realization of the tasks according to the established interdependencies.