Research

Research Interests

Visual data mining and visual analytics

text visualization, applications on visual data exploration and explainable AI

Machine learning

applications on data classification, clustering, active learning, data annotation

Natural language processing

text retrieval, text classification and clustering

Computer vision

applications on image classification, image preprocessing and pattern recognition on images using deep learning


If you are interested in applying for a undergraduate research project, with a possible scholarship, click here.

If you are interested in applying for a master's position under my supervision, click here.

Research Group

Grupo de Visualização e Inteligência Computacional (GVIC) on GitLab

Associated Research Labs

Laboratório de Raciocínio Automatizado (LARA)

Laboratório de Imagens, Sinais e Acústica (LISA)

DGP CNPq

Grupo de Inteligência Artificial (GIA)

Laboratório de Imagens, Sinais e Acústica (LISA)

Past Projects

KnEDLe - Knowledge Extraction from Documents of LEgal content

[2020-2023] Official publications from government gazettes (such as the Diário Oficial da União (DOU) and the Diário Oficial do Distrito Federal (DODF)) are valuable information sources related to the government actions on public civil services. Particularly regarding DODF, the high amount of publications that were issued over the past years make the analysis of these documents unfeasible when performed manually by auditors and other specialists. This scenario is appropriate to investigate computational approaches based on natural language processing, machine learning and visualization aiming at obtaining implicit and useful knowledge from raw texts associated to the official publications, which are mostly available in PDF files. The goal is to develop tools and techniques for recognizing important entities in order to obtain structured representation of the publications which can lead to other applications based on information retrieval on DODF editions. The proposed solutions can enhance the government's transparency and assist auditing tasks concerning the proper use of public financial resources.

Funding: Fundação de Apoio a Pesquisa do Distrito Federal (FAPDF).
Principal Investigators: Dr. Thiago de Paulo Faleiros (CIC/UnB) / Dr. Teófilo Emídio de Campos (CIC/UnB).
Contributors: Dr. Luis Paulo Faina Garcia (CIC/UnB), Dr. Vinícius R. P. Borges (CIC/UnB), Dr. Marcelo Grandi Mandelli (CIC/UnB), Carolina Alves Okimoto (CIC/UnB), Dr. Andrei Lima Queiroz (CIC/UnB), Dr. Ricardo Marcondes Marcacini (ICMC/USP).
Host: Universidade de Brasília (UnB).


Visual exploration of feature spaces to support green algae taxonomic classification

[2012-2016] This Ph.D. project has the objective of employing Information Visualization and Multidimensional Visualization techniques to support biologists in tasks involved in the taxonomic classification of fresh water green algae. Taxonomical classification of such organisms is problematic due to their feature variations. Similar features may be identify multiple algae species, which introduces errors and inconsistencies in manual classification. It is expected that employing appropriate multidimensional data visualization techniques can collaborate to improve the quality of the overall process, minimizing drawbacks and offering an improved alternative over manual classification. Features currently employed for classification include shape descriptors manually extracted from microscopy images of the algae and other taxonomical keys. In order to improve the classification process in terms of efficiency, reliability and robustness, we shall investigate and develop automatic shape-based feature extraction from images, and integrate tehm with other available features. The proposed solutions shall be initially evaluated comparing the classification results obtained with the proposed methods with those obtained with standard processes currently employed by the biology experts who collaborate in this project.

Funding: São Paulo Research Foundation (FAPESP).
Principal Investigator: Profa. Dra. Maria Cristina Ferreira de Oliveira.
Contributor: Prof. Dr. Bernd Hamann (University of California, Davis - UC Davis).
Host: Instituto de Ciências Matemáticas e de Computação (ICMC), Universidade de São Paulo, São Carlos (USP São Carlos).



Using variational methods for segmenting images by means of parametric and non-parametric approaches

[2010-2011] The goal of this project is to explore and to develop variational methods for segmenting images of different natures. A variational method is defined by an energy functional, which needs to be minimized in relation to its constituent function. However, as we are dealing with discrete approaches for computing that solution, the numerical solution of the associated Partial Differential Equation (PDE) needs to be computed. One main issue is to investigate numerical schemes for solving PDE. Another relevant task refers in setting the particular characteristics of the energy functional, as the statistical representation of image regions, the definition of a supervised or an unsupervised approach and the nature of images that are been handled (noisy, medical, real-world etc).

Funding: Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq).
Principal Investigator: Profa. Dra. Celia A. Zorzo Barcelos
Host: Universidade Federal de Uberlândia (UFU)



Arquitetura de Desenvolvimento para a Produção de Software de apoio às Atividades de Saúde PROCAD - CAPES

[2009-2012] The goal of this project is to develop technology and specialize personnel to explore operations of searching and similarity comparison in software development tools. Specifically, such softwares are modeled to run as a content-based retrieval system of medical images and its associated applications to the health field.

Funding: Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES).
Principal Investigators: Profa. Dra. Agma Juci Machado Traina (ICMC/USP São Carlos), Prof. Dr. Caetano Traina Junior (ICMC/USP São Carlos), Prof. Dra. Denise Guliato (UFU)



Sistema de Apoio ao Diagnóstico de Câncer de Mama Integrando Sistema de Pesquisa por Conteúdo e Tutorial em Câncer de Mama

[2007-2009] This project proposes the development of two methods to breast segmentation in digital mammographic images. The first method is relative with the detection of external breast contour and the latter describes the pectoral muscle extraction procedure in Middle Lateral Oblique (MLO) views. Both methods were tested in images randomly selected from the public database Digital Database for Screening Mammography (DDSM). The results show that the proposed methods are very promising and effective in extration of objects that not belongs at the breast structure and can contribute positively in detection and classification of suspect regions. (CNPq)

Funding: Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq).
Principal Investigator: Profa. Dra. Denise Guliato
Universidade Federal de Uberlândia