An essential task to ensure the identification of vehicles and safety on roads and private spaces, Automatic License Plate Recognition (ALPR) is the research subject of Gabriel Gonçalves, a Ph.D. student at the Sense lab, who will present his work at the Conference on Graphics, Patterns and Images (SIBGRAPI). In its 31^st edition, SIBGRAPI is the prenier conference on computer vision and graphics in Latin America and will be held from October 29^th to November 1^st at the Bourbon Cataratas Convention & Spa Resort, in Foz do Iguaçu, Paraná.

With the increasing number of surveillance cameras available in urban areas, video-based traffic analysis can provide useful insights for the transportation segment. One such analysis is Automatic License Plate Recognition. Previously, this was usually performed as a series of tasks: vehicle detection, license place detection, character localization, and character recognition. Gonçalves’ work, “Real-time Automatic License Plate Recognition Through Deep Multi-Task Networks”, proposes an alternative approach: employing only two neural networks to perform the entire license plate recognition pipeline.

“Besides the gain on robustness due to the use of deep learning techniques, this also improves execution time as the approach is able to run in real-time even without a powerful computer. In practice, this means that any private or public company can use our approach without much investment on equipment”, explains Gonçalves.

In addition, acording to experimental results, the approach proposed is able to achieve state-of-the-art results in the Sense-SegPlate dataset, reaching improvements of up to 1.4 percentage point when compared to the best baseline.

Part of the research that will be presented by Gabriel was developed in partnership with the tracking and telemetry company Maxtrack through the Brazilian Company of Research and Innovation (EMBRAPII).

The researcher

With experience mainly in visual pattern recognition and deep learning, Gonçalves holds a degree in Computer Science from the Federal University of Ouro Preto (UFOP). He obtained his Master’s degree in Computer Science from the Federal University of Minas Gerais (UFMG) with the thesis “License Plate Recognition based on Temporal Redundancy”. He has worked as a research fellow at the Research Development Foundation (FUNDEP) in partnership with Hewlett-Packard and Maxtrack.

The publication

Led by Gonçalves, the work had substantial contribution from the Sense lab Master’s student Matheus Alves Diniz, who helped in the development of methodology and experiments. The project was done in collaboration with UFPR researchers Rayson Laroca and David Menotti. Professor William Schwartz, coordinator of the Smart Sense Laboratory, advised it. The publication can be accessed through this link.

The conference

The Conference on Graphics, Patterns and Images is a leading annual event combining contributions from four major subjects related to image computing: computer graphics & vision, pattern recognition and image processing. This year SIBGRAPI will be co-located with two other major events: SBGames 2018, the Symposium on Computer Games and Digital Entertainment, and SVR 2018, the Symposium on Virtual and Augmented Reality.