Alunos do Mestrado e professores do Campus Serra têm trabalhos aceitos no WVC 2016

• Imprimir

Os alunos do Mestrado em Engenharia de Controle e Automação, Ricardo Koscky, Eduardo Pereira, Warley Mendes e Vinicius da Motta e os professores da Coordenadoria de Auotmação, Luiz Alberto Pinto, Flávio Garcia Pereira, Karin Komati, Daniel Cavalieri e Jefferson Andrade tiveram seus trabalhos aceitos no Workshop de Visão Computacional (WVC).

O Workshop de Visão Computacional tem por objetivos reunir pesquisadores, estudantes de pós-graduação e de graduação, para discutir o estado da arte na área. Este ano, o WVC será em Campo Grande, MS, de 09 à 11 de novembro. Confira abaixo os trabalhos aprovados:
1) "Audio Transient Detection - Applying Computer Vision over Spectrograms" - apresentação oral
* abstract: Transient detection in audio material is an important step towards music information retrieval (MIR) as it enables onsets localization over time and the determination of tempo and other song’s metrics. Complementing methods already available, this paper introduces a strategy of feature extraction directly from spectrograms image analysis, using computer vision techniques (e.g.: filtering). This strategy was tested over a well-known audio dataset and proved to be an easy and quick way to detect not only the transients’ timing, but additionally, also their strength.
* Autores: Ricardo Koscky (aluno), Luiz Alberto Pinto, Flávio Garcia Pereira, Karin Komati 

2) "Leaf classification using classifier fusion" 
* abstract: The purpose of this work is to use the fusion of classifiers SVM (Support Vector Machine) and PLS (Partial Least Square) to get results above 95% of assertiveness in classification of 1907 pictures of leaves in 32 different plant species, surpassing this way the results obtained by other methods used for the same purpose.
* Autores: Eduardo Pereira (aluno) , Warley Mendes (aluno), Karin Komati, Daniel Cavalieri, Luiz Alberto Pinto , Flávio Garcia Pereira 

3) "Detection and Tracking of Pedestrians and Cyclists on a beach promenade - Naive Method" 
* abstract: Every year, more than 40 thousand deaths are caused by traffic accidents, and modes of transit with highest fatality hates are pedestrian and cyclist. It is paramount that new technologies provide support to prevent and detect such accidents in a timely manner. One such technology is the use of digital image processing to identify and track pedestrians and cyclists. This work presents an low cost approach of detecting and tracking cyclists and pedestrians, built of individual steps that apply, sequentially, different processing techniques. We then tested a prototype based on our technique on a real-world video feed. Our method displayed an improvement of more than 700% on the number of objects detected over a standard implementation of object detection provided by MATLAB. Our prototype will lead to the development of new, more powerful tools, that can be used to prevent and detect traffic accidents involving pedestrians and cyclists.
* Autores: VINICIUS DA MOTTA (aluno), Karin Komati , Jefferson Andrade.

Para mais informações acesse: http://wvc2016.weebly.com/