14/11/2019

Visgraf integrates a scientific research project of the BRICS

Drone do IME faz sobrevoo

Imagine the following scene: a drone, operated by Brazilian researchers, flies over a plot of land in the South Zone of Rio de Janeiro. The goal is to record images and create a detailed database of that area. All the information collected will be sent to another research group.

In this second phase, taking place in China, professionals will work in the field of computer graphics, with a particular focus on geometric processing and modeling techniques. Even without having set foot in Brazil, the researchers are able to reconstruct the captured objects and images in 3D. Following this, the calculations regarding the Brazilian region will be sent to Russia, where scientists are developing new methods of machine learning, deep neural networks, computer vision, and remote sensing processing.

To integrate the entire process, from Brazil to Russia, there is an intense exchange of emails, Skype conferences, and data in the cloud. All of this may seem like science fiction, but it is part of the daily routine of researchers at Visgraf (the Computer Graphics Laboratory of IMPA), the Military Engineering Institute (IME), the Skolkovo Institute of Science and Technology, and Zhejiang University. The BRICS research group transcends geographical boundaries to advance the state of the art. The researchers are part of the "Three-Dimensional Reconstruction by Machine Learning" project and also have the support of the Courant Institute for Mathematical Sciences at New York University.

Parte da equipe de pesquisadores brasileiros

The research centers participating in the project are internationally recognized in their respective fields and possess complementary expertise across the numerous areas in which they operate. Visgraf contributes its experience in geometric modeling, three-dimensional reconstruction, and the generation of 3D external scenes captured with UAVs (unmanned aerial vehicles). Researchers at IME develop the equipment used for image capture. In this type of mapping, the use of more than one drone is common. The machines are connected and communicate via the internet to map as much available information as possible.

The group was officially established in 2018 and is expected to continue its work until 2021. During this period, researchers intend to generate applications for three-dimensional scene reconstruction, based on real remote sensing data. For Luiz Velho, lead researcher of Visgraf, the project "Three-Dimensional Reconstruction by Machine Learning" is ambitious. "The area of machine learning has gained a lot of momentum in recent years. The great advance was the application of machine learning to deep networks for images. But working with images is much easier than working with three-dimensional data."

The group's biggest challenge is precisely trying to make this upgrade between machine perception systems so that data processing happens not only in two dimensions, but also in 3D. The technique can be applied to improve, for example, the analysis of human movements, architecture, facial recognition, and biomedical data processing, such as magnetic resonance imaging.

Luiz Velho believes that the BRICS project is also very important in encouraging more experienced researchers to continue their work. For him, just as the BRICS Young Scientists Forum inspires those starting research, the exchange and sharing of experiences that have been happening with researchers from China and Russia are some of the factors that encourage those who have been in the scientific field for several years.