Randomized neural network based signature for dynamic texture classification

Autor(es)

de Mesquita Sa Junior, Jarbas Joaci ^[1] ; Ribas, Lucas Correia ^[2] ; Bruno, Odemir Martinez ^[3]

Afiliação do(s) autor(es)

^[1] Univ Fed Ceara, Programa Posgrad Engn Eletr & Comp, Curso Engn Computacao, Campus Sobral, Rua Coronet Estanislau Frota 563, BR-62010560 Sobral, Ceara - Brazil

^[2] Univ Sao Paulo, Inst Math & Comp Sci, Ave Trabalhador Stio Carlene 400, BR-13566590 Sao Carlos, SP - Brazil

^[3] Univ Sao Paulo, Sao Carlos Inst Phys, POB 369, BR-13560970 Sao Carlos, SP - Brazil

• Tipo de documento: Artigo Científico
• Fonte: EXPERT SYSTEMS WITH APPLICATIONS; v. 135, p. 194-200, NOV 30 2019.
• Citações Web of Science: 1

Resumo

Dynamic texture analysis has been the focus of intensive research in recent years. Thus, this paper presents an innovative and highly discriminative dynamic texture analysis method, whose signature is composed of the weights of the output layer of a randomized neural network after a training procedure. This training is performed by using the pixels of slices of each orthogonal plane of the video (XY, YT, and XT) as input feature vectors and corresponding output labels. The obtained video signature provided an accuracy of 97.05%, 98.54%, 97.74% and 96.51% on the UCLA-50 classes, UCLA-9 classes, UCLA-8 classes and Dyntex++, respectively. These results, when compared to other dynamic texture analysis methods, demonstrate that our descriptors are very effective and that our proposed approach can contribute significantly to the field of dynamic texture analysis. (C) 2019 Elsevier Ltd. All rights reserved. (AU)

• Processo FAPESP: 16/23763-8 - Modelagem e análise de redes complexas para visão computacional
• Beneficiário: Lucas Correia Ribas
• Linha de fomento: Bolsas no Brasil - Doutorado
• Processo FAPESP: 14/08026-1 - Visão artificial e reconhecimento de padrões aplicados em plasticidade vegetal
• Beneficiário: Odemir Martinez Bruno
• Linha de fomento: Auxílio à Pesquisa - Regular
• Processo FAPESP: 16/18809-9 - Deep learning e redes complexas aplicados em visão computacional
• Beneficiário: Odemir Martinez Bruno
• Linha de fomento: Auxílio à Pesquisa - Parceria para Inovação Tecnológica - PITE