Research Interests:
Research Interests:
There are some benefits in using periocular biometric traits for individual identification. This work describes the use of convolutional neural network Neocognitron, in this novel application, in individual recognition using periocular... more
There are some benefits in using periocular biometric traits for individual identification. This work describes the use of convolutional neural network Neocognitron, in this novel application, in individual recognition using periocular region images. Besides, it is used the competitive learning using the extreme points of lines detected in the preprocessing of the input images as winner positions. It was used Carnegie Mellon University - Pose, Illumination, and Expression Database (CMU-PIE), with 41,368 images of 68 persons. From these images, 57 \(\times \) 57 periocular images were obtained as training and test samples. The experiments indicate results in the Kappa index of 0.89, for periocular images, and 0.91 for complete face images.
Research Interests:
Research Interests:
The biometric periocular trait refers to the face region in the vicinity of the eye, including the eyelids, eyelashes and eyebrows. The periocular region has emerged as a promising trait for unconstrained biometrics, due to recent... more
The biometric periocular trait refers to the face region in the vicinity of the eye, including the eyelids, eyelashes and eyebrows. The periocular region has emerged as a promising trait for unconstrained biometrics, due to recent advances of convolutional neural networks and the demand for robust face or iris recognition systems. The periocular region can offer global information about the eye shape, and about the texture of the iris, sclera and skin around the eyes. However, periocular biometrics is a relatively new area of research. Thus, it's important to understand the uniqueness and stability of this trait, taking into account the best accuracies obtained by deep learning methods applied on biometric image recognition. In this work, we investigate if changes in the periocular region, caused by facial expressions, affect the recognition accuracy. We apply an existing pretrained CNN architecture, called MobileNet, to the task of periocular recognition. The periocular images used in the experiments were extracted from the Extended Cohn-Kanade expression database. The best results were obtained when the network was tested with similar samples to those contained in the training set. We concluded that the CNN is sensitive to changes in the facial expressions and samples of all expressions are required for training aiming the best accuracy.
Research Interests:
Recent research in convolutional neural network (CNN) has provided a variety of new architectures for deep learning. One interesting new architecture is the local binary convolutional neural network (LBCNN), which has shown to provide... more
Recent research in convolutional neural network (CNN) has provided a variety of new architectures for deep learning. One interesting new architecture is the local binary convolutional neural network (LBCNN), which has shown to provide significant reduction in the number of parameters to be learned at training. In this paper, we study the influence of network parameters in the scenario of face recognition, comparing LBCNN against other famous networks available in the literature in terms of sensibility and processing time. In our study, we also propose a pre-processing step on images to increase the accuracy of the model, besides investigating its behaviour with noisy images. Our experiments are carried on the Chokepoint dataset, whose face subimages were collected from video frames under real-world surveillance conditions, including variations in terms of illumination, sharpness, pose, and misalignment due to automatic face detection. The conclusion is that by using the Laplacian step and a reduced amount of LBC modules, it is possible to train LBCNN more quickly and with improved accuracy. In addition, it was found that LBCNN is very sensitive to noise and better results can be achieved when noisy images are inserted in the training set.
Research Interests:
Research Interests:
ABSTRACT The manual selection of linear and nonlinear operators for producing image filters is not a trivial task in practice, so new proposals that can automatically improve and speed up the process can be of great help. This paper... more
ABSTRACT The manual selection of linear and nonlinear operators for producing image filters is not a trivial task in practice, so new proposals that can automatically improve and speed up the process can be of great help. This paper presents a new proposal for constructing image filters using an evolutionary programming approach, which has been implemented as the IFbyGP software. IFbyGP employs a variation of the Genetic Programming algorithm GP and can be applied to binary and gray level image processing. A solution to an image processing problem is represented by IFbyGP as a set of morphological, convolution and logical operators. The method has a wide range of applications, encompassing pattern recognition, emulation filters, edge detection, and image segmentation. The algorithm works with a training set consisting of input images, goal images, and a basic set of instructions supplied by the user, which would be suitable for a given application. By making the choice of operators and operands involved in the process more flexible, IFbyGP searches for the most efficient operator sequence for a given image processing application. Results obtained so far are encouraging and they stress the feasibility of the proposal implemented by IFbyGP. Also, the basic language used by IFbyGP makes its solutions suitable to be directly used for hardware control, in a context of evolutionary hardware. Although the proposal implemented by IFbyGP is general enough for dealing with binary, gray level and color images, only applications using the first two are considered in this paper; as it will become clear in the text, IFbyGP aims at the direct use of induced sequences of operations by hardware devices. Several application examples discussing and comparing IFbyGP results with those obtained by other methods available in the literature are presented and discussed.
Research Interests:
Research Interests:
Microelectrode Array (MEA) have been used for electrophysiological signal recording of the activity of neuron cells. These signals are used to the study of the neuronal network dynamics in neuroscience; as well as for the drug tests in... more
Microelectrode Array (MEA) have been used for electrophysiological signal recording of the activity of neuron cells. These signals are used to the study of the neuronal network dynamics in neuroscience; as well as for the drug tests in pharmacological applications, between others. The majority of MEA recording systems in the world, needs several amplifier blocks (one per channel). These amplifier blocks, always in a great number, result in a very complex and high cost implementation. It is proposed a novel data acquisition system for a MEA of 60 microelectrodes, in this paper. The prototype tests showed that the recording of microelectrode signals at mV level, using a single headstage pre-amplifier, filter, amplifier, analog-to-digital converter, and USB converter, are feasible.
Research Interests:
Research Interests:
Research Interests:
ABSTRACT Mathematical morphology is a formalism largely used in image processing for implementing many different tasks. Several operators that support the formalism have also been successfully used for inducing data clusters.... more
ABSTRACT Mathematical morphology is a formalism largely used in image processing for implementing many different tasks. Several operators that support the formalism have also been successfully used for inducing data clusters. Particularly, the Binary Morphology Clustering Algorithm (BMCA) is one of such inductive methods which, given a set of input patterns and morphological operators, produces clusters of patterns as output. BMCA results, however, are dependent on suitable user-defined values for the set of parameters the algorithm employs namely, the resolution of its initial discretization process, the threshold associated with a distance metric, the threshold associated with region density and the structuring element embedded in morphological operators. This paper proposes a combined approach where an evolutionary algorithm is employed for searching suitable parameter values for BMCA aiming at producing more efficient results as far as the clustering process is concerned. The proposal was implemented as the system BMCAbyGA, used in several successful clustering experiments described in the final part of the paper. BMCAbyGA has been applied to a Cartesian Genetic Programming approach for the automatic construction of image Alters in hardware.
Research Interests:
Research Interests:
Research Interests:
... The network was visualized by using the free software Gephi (http://gephi.org/). The layout was obtained through the algorithm “Force At-las” in Gephi. This is a force-based algorithm whose principle is to attract linked ver-tices and... more
... The network was visualized by using the free software Gephi (http://gephi.org/). The layout was obtained through the algorithm “Force At-las” in Gephi. This is a force-based algorithm whose principle is to attract linked ver-tices and push apart non-linked vertices. ...
Research Interests:
Research Interests:
ABSTRACT
Research Interests:
ABSTRACT This work studies the evolution of neuronal network of in vitro cultures of neurons, in Multi-Electrode Array (MEA), using a multivariate autoregressive method and a partial directed coherence technique to obtain the causal... more
ABSTRACT This work studies the evolution of neuronal network of in vitro cultures of neurons, in Multi-Electrode Array (MEA), using a multivariate autoregressive method and a partial directed coherence technique to obtain the causal relations between electrodes. The electrophysiological data of the hippocampal neurons of 18 days old Wistar rat embryos, were registered at Genoa University, with 3 days interval, from 25 DIV (Days In Vitro) to 46 DIV. The analysis of the obtained connectivity results, using the above mentioned statistical methods, was performed evaluating complex network properties, considering the MEA electrodes as nodes. The conclusion is that the described method is adequate to analyze the evolution of the cultured neuronal network.
Research Interests:
Mathematical morphology supplies powerful tools for low-level image analysis. Many applications in computer vision require dedicatedhardwarefor real-time execution. The design of morphological operators for a given application is not a... more
Mathematical morphology supplies powerful tools for low-level image analysis. Many applications in computer vision require dedicatedhardwarefor real-time execution. The design of morphological operators for a given application is not a trivial one. Genetic programming is a branch of evolutionary computing, and it is consolidating as a promising method for applications of digital image processing. The main objective of genetic programming is to discover how computers can learn to solve problems without being programmed for that. In this paper, the development of an original reconfigurable architecture using logical, arithmetic, and morphological instructions generated automatically by a genetic programming approach is presented. The developed architecture is based on FPGAs and has among the possible applications, automatic image filtering, pattern recognition and emulation of unknown filter. Binary, gray, and color image practical applications using the developed architecture are pre...
Research Interests:
Research Interests: Computer Science, Parallel Computing, Image Processing, Computed Tomography, Digital Signal Processing, and 10 moreParallel Processing, Image Reconstruction, Parallel Imaging, Performance Improvement, Sample Size, Message Passing Interface, Electrical And Electronic Engineering, Multiprocessing, Iterative Reconstruction, and image resolution
Resumo Esse artigo descreve uma aplicação de redes neurais para visualização de relevos naturais como composições musicais. Durante a fase de treinamento, a rede neural aprende certos aspectos da estrutura musical através de exemplos de... more
Resumo Esse artigo descreve uma aplicação de redes neurais para visualização de relevos naturais como composições musicais. Durante a fase de treinamento, a rede neural aprende certos aspectos da estrutura musical através de exemplos de compassos extraídos de melodias do conjunto de treinamento. As melodias escolhidas para o treinamento são melodias folclóricas, por serem simples e monofônicas. Durante a composição, o sistema utiliza o repertório de compassos usados no treinamento para compor novas melodias, ...
Research Interests:
Research Interests:
Matriz de Microeletrodos (microelectode array, MEA) é um dispositivo usado para cultura de neurônios in-vitro e registro de sinais eletrofisiológicos. A análise desses sinais é útil para o estudo da dinâmica de redes neuronais, teste de... more
Matriz de Microeletrodos (microelectode array, MEA) é um dispositivo usado para cultura de neurônios in-vitro e registro de sinais eletrofisiológicos. A análise desses sinais é útil para o estudo da dinâmica de redes neuronais, teste de drogas farmacológicas, entre outras aplicações. Os sinais elétricos capturados pela MEA são de baixa amplitude e requer sistemas de amplificação especiais. Neste artigo, é descrito um sistema de amplificação para MEA, com o uso do fenômeno de Ressonância Estocástica (RE), que faz uso benéfico do ruído, para detectar e amplificar sinais de microeletrodos de pequena amplitude. Um protótipo de um canal desse sistema foi simulado e testado com sucesso.
Este trabalho apresenta um sistema multiprocessador voltado ao processamento de imagens, utilizando processadores digitais de sinais TMS320C30 da Texas Instruments, incluindo os elementos de hardware e software funcionamento. Alguns... more
Este trabalho apresenta um sistema multiprocessador voltado ao processamento de imagens, utilizando processadores digitais de sinais TMS320C30 da Texas Instruments, incluindo os elementos de hardware e software funcionamento. Alguns aspectos de apresentados através de resultados necessários para o análise de desempenho obtidos por simulação sistema executando a transformada de Fourier bidimensional.
Research Interests:
Este trabalho apresenta a análise de desempenho de uma arquitetura multiprocessadora composta de Processadores Digitais de Sinais (DSP) da Texas lnstruments TMS320C40 operando sob diferentes algoritmos de comunicação entre processos. Esta... more
Este trabalho apresenta a análise de desempenho de uma arquitetura multiprocessadora composta de Processadores Digitais de Sinais (DSP) da Texas lnstruments TMS320C40 operando sob diferentes algoritmos de comunicação entre processos. Esta arquitetura foi organizada para a reconstrução tridimensional de cortes tomográficos e testes foram realizados com dados obtidos de um minitomógrafo® de solos da Embrapa Instrumentação Agropecuária. Foram utilizados até 4 processadores TMS320C40 acoplados a módulos TIM-40 em uma placa HEPC2E. No desenvolvimento do software utilizou-se uma combinação das ferramentas Code Composer da Texas lnstruments, da linguagem C Paralela da 3L e do Borland Builder C++. As tarefas de reconstrução 2D e 3D foram distribuídas entre os processadores e utilizou-se a memória cache dos processadores para diminuição das taxa de perda e aumento do poder de processamento e da velocidade do sistema. Também utilizou-se algoritmos de particionamento das matrizes de reconstruç...
Research Interests:
Esse trabalho apresenta um sistema para reconstrução de imagens tomográficas usando uma plataforma paralela dedicada para o minitomógrafo de solos da Embrapa Instrumentação Agropecuária. Foram utilizados dois processadores TMS320C40... more
Esse trabalho apresenta um sistema para reconstrução de imagens tomográficas usando uma plataforma paralela dedicada para o minitomógrafo de solos da Embrapa Instrumentação Agropecuária. Foram utilizados dois processadores TMS320C40 acoplados a módulos TIM-40 em uma placa HEPC2E. No desenvolvimento do software utilizou-se uma combinação das ferramentas Code Composer da Texas lnstruments, da linguagem C Paralela da 3L e do Borland Builder C++. As tarefas de reconstrução 20 e 3D foram distribuídos entre os processadores e utilizou-se a memória cache dos processadores para diminuição das taxa de perda e aumentar tanto o poder de processamento quanto à velocidade do sistema. Resultados com dados de amostras tomográficas reais utilizando o método de retroprojeção mostram boa performance, com uma redução de aproximadamente 80% do tempo de reconstrução demandado em plataforma convencional com um microprocessador de 200 MHz.
Research Interests:
Hyperspectral images and remote sensing are important for many applications. A problem in the use of these images is the high volume of data to be processed, stored and transferred. Dimensionality reduction techniques can be used to... more
Hyperspectral images and remote sensing are important for many applications. A problem in the use of these images is the high volume of data to be processed, stored and transferred. Dimensionality reduction techniques can be used to reduce the volume of data. In this paper, an approach to band selection based on clustering algorithms is presented. This approach allows to reduce the volume of data. The proposed structure is based on Fuzzy C-Means (or K-Means) and NWHFC algorithms. New attributes in relation to other studies in the literature, such as kurtosis and low correlation, are also considered. A comparison of the results of the approach using the Fuzzy C-Means and K-Means with different attributes is performed. The use of both algorithms show similar good results but, particularly when used attributes variance and kurtosis in the clustering process, however applicable in hyperspectral images.
Research Interests:
This work is concerned to an analysis of the use of alternative BDA Correlation Systems implementations. The first one is the FPGA (Field P rogrammable Gate Array) system, using commercial platforms of reconfigurable chips, in th is case... more
This work is concerned to an analysis of the use of alternative BDA Correlation Systems implementations. The first one is the FPGA (Field P rogrammable Gate Array) system, using commercial platforms of reconfigurable chips, in th is case a Virtex-II Pro (XC2VP30) of Xilinx Company. The second alternative is the softw are correlation. In this case, a software development platform, explicitly the GPU (Graphic P rocessing Unit) processor, with multiple cores driven by very high memory bandwidth, is anal yzed. The results showed the viability of this type of device as a massively parallel data pr ocessing tool, and that smaller the granularity of the parallel processing, and the ind ependence of the processing, better is its performance.
Research Interests:
The aim of this paper is to evaluate the performance of Transfer Learning techniques applied in Convolucional Neural Networks for biometric periocular classification. Two aspects of Transfer Learning were evaluated: the technique known as... more
The aim of this paper is to evaluate the performance of Transfer Learning techniques applied in Convolucional Neural Networks for biometric periocular classification. Two aspects of Transfer Learning were evaluated: the technique known as Fine Tuning and the technique known as Feature Extraction. Two CNN architectures were evaluated, the AlexNet and the VGG-16, and two image databases were used. These two databases have different characteristics regarding the method of acquisition, the amount of classes, the class balancing, and the number of elements in each class. Three experiments were conducted to evaluate the performance of the CNNs. In the first experiment we measured the Feature Extraction accuracy, and in the second one we evaluated the Fine Tuning performance. In the third experiment, we used the AlexNet for Fine Tuning in one database, and then, the FC7 layer of this trained CNN was used for Feature Extraction in the other database. We concluded that the data quality (the ...
Research Interests:
This paper presents the PVA-BDA project (Processing, Visualization and Analysis in ParallelEnvironment of the BDA Data) that has been developed for processing of solar images that will be captured bythe BDA (Brazilian Decimetric Array), a... more
This paper presents the PVA-BDA project (Processing, Visualization and Analysis in ParallelEnvironment of the BDA Data) that has been developed for processing of solar images that will be captured bythe BDA (Brazilian Decimetric Array), a radio telescope under development at the National Institute for SpaceResearch (INPE). In a joint effort between the Department of Computer Science at Federal University of SãoCarlos (DC/UFSCar), the Astrophysics Division (DAS) and Associated Laboratory for Computing and AppliedMathematics (LAC) at INPE, a high performance parallel system is being developed with capacity to supportrealistic applications, involving a reasonable amount of parallel processing, in order to carry out the processing,visualization and analysis of solar images captured by BDA, in real time. The aim is to create the conditions forstarting a study of the solar weather forecast. The forecast of solar explosions are important as they may causeserious perturbations in terrestria...
Research Interests:
Page 1. A 3D SCANNING SYSTEM BASED ON LASER TRIANGULATION AND VARIABLE FIELD OF VIEW João Guilherme DM França1, Mário A. Gazziro1, Alessandro N. Ide2, José H. Saito1 {jgfranca, mariog}@dc.ufscar.br, noriaki@dist.unige.it,... more
Page 1. A 3D SCANNING SYSTEM BASED ON LASER TRIANGULATION AND VARIABLE FIELD OF VIEW João Guilherme DM França1, Mário A. Gazziro1, Alessandro N. Ide2, José H. Saito1 {jgfranca, mariog}@dc.ufscar.br, noriaki@dist.unige.it, saito@dc.ufscar.br ...
Research Interests:
Research Interests:
Research Interests:
Research Interests: Computer Science, Computer Architecture, Parallel Computing, Face Recognition, Parallel Processing, and 11 moreNeural Network, Compute Unified Device Architecture NVIDIA CUDA, Graphics, CUDA, High performance, Load Balance, Data Processing, High performance computer, Artificial Neural Network, graphic processing unit (GPU), and Graphics processing unit
We are carrying out a development of a whole system for processing solar images in a high performance parallel system. The main objective is to create the initial conditions for studying and forecasting the solar explosions in real time.... more
We are carrying out a development of a whole system for processing solar images in a high performance parallel system. The main objective is to create the initial conditions for studying and forecasting the solar explosions in real time. Due to the high computational costs involved in the processing, visualization and analysis of a great amount of solar images, a high performance computer system becomes necessary to carry out the forecast of solar explosions. As a joint effort between the Department of Computer Science at Federal University of São Carlos (UFSCar), the Astrophysics Division (DAS) and Associated Laboratory for Computing and Applied Mathematics (LAC) at National Institute for Space Research - INPE, a high performance parallel system was developed with capacity to support realistic applications, involving a reasonable amount of parallel processing. The forecast of solar explosions is important as they may cause serious perturbations in terrestrial communication systems....