Ongoing research projects

The AdapLin Tool is an adaptive, interactive, assisting tool for roads vector mapping. It is based on dynamic programming and was developed as a plugin for the QGIS software. AdapLin is intended to optimize the vector mapping of roads, based on features present on a georeferenced image.

Funding: Scientific initiation scholarships program of Rio de Janeiro State University

This project is dedicated to the development of technology to support automatic athlete tracking in video sequences.

The project aims at contributing to the smart cities field through the development of technology able to support the automatic quality control of Brazilian traffic signs underlain on computer vision and machine learning.

Investigation in collaboration with LVC (Computer Vision Lab) at Pontifical Catholic University of Rio de Janeiro and Institute of Photogrammetry and GeoInformation at Hannover University about the usage of deep architectures in the context of semantic labeling/segmentation. Its main concern is to evaluate different strategies to handle the insufficient availability of labelled training data, towards fully exploiting the potential of deep architectures for remote sensing image analysis.

Funding: CAPES and DAAD through the PROBRAL interchange program.

Segmentation is among the most costly parts of image processing workflow. This project encompasses several researches aiming at speeding up image segmentation. The main techniques employed for that purpose are parallel programming and high performance computing by using multi-core processors, GPUs and distributed systems.

Funding: Local internship program of Rio de Janeiro State University