Seminário com o Eduardo Luz, dia 02/06/17, as 13:30 no DECOM.

Seminário com o Eduardo Luz, dia 02/06/17, as 13:30 na Sala de Seminários do DECOM.


"Exploring deep learning representations for biometric multimodal systems"


Usually, in the deep learning community, it is claimed that generalized representations that yield outstanding performance/effectiveness require an enormous amount of data for learning, which directly affect multimodal biometric applications due to lack of available data. However, recent works combining data augmentation and transfer learn- ing have surmounted such volume of data constraints yielding promising deep learning approaches in different scenarios where data is not so abundant. In this direction, data augmentation and transfer learning techniques are explored here in biometric scenarios with limited data. Results showed that deep learning could be a powerful tool for improving multimodal systems, especially on uncontrolled and non-cooperative databases.

Departamento de Computação  |  ICEB  |  Universidade Federal de Ouro Preto
Campus Universitário Morro do Cruzeiro  |  CEP 35400-000  |  Ouro Preto - MG, Brasil
Telefone: +55 31 3559-1692  |