Resumo – Publicações

Learning graph representation with randomized neural network for dynamic texture classification.
RIBAS, Lucas Correia; SÁ JÚNIOR, Jarbas Joaci de Mesquita; MANZANERA, Antoine; BRUNO, Odemir Martinez.
Abstract: Dynamic textures (DTs) are pseudo periodic data on a space×time support, that can representmany natural phenomena captured from video footages. Their modeling and recognition are usefulin many applications of computer vision. This paper presents an approach for DT analysis combininga graph-based description from the Complex Network framework, and a learned representation fromthe Randomized Neural Network (RNN) model. First, a directed space×time graph modeling withonly one parameter (radius) is used to represent both the motion and the appearance of the DT. Then,instead of using classical graph measures as features, the DT descriptor is learned using a RNN, thatis trained to predict the gray level of pixels from local topological measures of the graph. The weightvectorofthe outputlayerof theRNNformsthe descriptor.Severalstructures areexperimentedfor theRNNs, resulting in networks with final characteristics of a single hidden layer of 4, 24, or 29 neurons,and input layers of sizes 4 or 10, meaning 6 different RNNs. Experimental results on DT recognitionconducted on Dyntex++ and UCLA datasets show a high discriminatory power of our descriptor,providing an accuracy of 99.92%, 98.19%, 98.94% and 95.03% on the UCLA-50, UCLA-9, UCLA-8 andDyntex++ databases, respectively. These results outperform various literature approaches, particularlyfor UCLA-50. More significantly, our method is competitive in terms of computational efficiency anddescriptor size. It is therefore a good option for real-time dynamic texture segmentation, as illustratedby experiments conducted on videos acquired from a moving boat.
Applied Soft Computing
v. 114, p. 108035-1-108035-14 - Ano: 2022
Fator de Impacto: 6,725
    @article={003053264,author = {RIBAS, Lucas Correia; SÁ JÚNIOR, Jarbas Joaci de Mesquita; MANZANERA, Antoine; BRUNO, Odemir Martinez.},title={Learning graph representation with randomized neural network for dynamic texture classification},journal={Applied Soft Computing},note={v. 114, p. 108035-1-108035-14},year={2022}}