Modelo baseado em deep learning para detecção de portas e escadas para auxiliar deficientes visuais

• Ano de defesa: 2018
• Autor(a) principal: Nogueira Junior, Rodney Sales
• Orientador(a): Manssour, Isabel Harb
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Pontifícia Universidade Católica do Rio Grande do Sul
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação
• Departamento: Escola Politécnica
• País: Brasil
• Palavras-chave em Português: Visão Computacional; Redes Neurais Convolucionais; Ambientes Internos; Deficientes Visuais
• Palavras-chave em Inglês: Computer Vision; Convolutional Neural Networks; Indoor Environments; Visually Impaired
• Área do conhecimento CNPq: CIENCIA DA COMPUTACAO::TEORIA DA COMPUTACAO
• Link de acesso: http://tede2.pucrs.br/tede2/handle/tede/9828

Due to a large number of visually impaired persons in the world, and with the advance of technology, the research interest in the development of different approaches to support the mobility of these persons has increased. In this context, the detection of doorsand stairs is an important research topic because it provides useful information that can aid in the mobility of these persons. In this work, we present a model to aid the visually impaired navigation in indoor environments. We found approaches that use computer vision techniques to identify corridors, obstacles, stairs, and doors through a literature review. However, few of them use recent techniques in computer vision and convolutional neural networks in their solutions. Thus, the presented model includes an experiment on convolutional neural networks to recognize and detect doors and stairs. Using the YOLO method, we present a model that detects not only different kinds of doors but also is capable of differentiating ascending and descending stairs, with FPS rates close to 30 and mAP above 90%.