Ampliação da base MINDS-Libras: um estudo de aplicação de técnicas de aumento sintético de dados e da inclusão de novos conjuntos de vídeos disponíveis na literatura.

Arthur Cesar da Silva Almeida; Tamires Martins Rezende; Sílvia Grasiella Moreira Almeida

doi:10.20906/CBA2022/3424

Arthur Cesar da Silva Almeida Departamento de Engenharia de Controle e Automação, Universidade Federal de Ouro Preto, Campus Universitário Morro do Cruzeiro, Bauxita, Ouro Preto Minas Gerais
Tamires Martins Rezende Machine Intelligence and Data Science (Minds) Laboratory, Universidade Federal de Minas Gerais, Av. Antônio Carlos 6627 - Escola de Engenharia - Bloco I, Sala 2200, Belo Horizonte, Minas Gerais; Fundação de Inovações Tecnológicas (FITec), Av Cristóvão Colombo, 485, 8º Andar – Belo Horizonte, MG
Sílvia Grasiella Moreira Almeida Instituto Federal de Minas Gerais (IFMG) - Campus Ouro Preto, Rua Pandiá Calógeras 898, Bauxita, Ouro Preto, Minas Gerais

DOI: https://doi.org/10.20906/CBA2022/3424

Keywords: Libras dataset, dataset expansion, Libras sign recognition

Abstract

This paper aim the study and sign Libras cataloging for further search for the MINDS- Libras dataset. Although there is progress in the availability of sign bases in Libras for use by the scientific community, as is the case with MINDS-Libras, it is still a challenge for a Libras sign recognition area to have access to a sample of signs for training computer systems that can recognize signals automatically. It intends to expand to a MINDS-Libras base of two: evaluating the generation of synthetic data and analyzing the large mass of data publicly formed during the remote activities that take place in the period of the Covid-19 pandemic. The initiative for this work is in the contribution that they intend to force for the Computational vision of recognition that impacts directly in the social and cultural areas for the emerged community.