%0 Conference Proceedings
%4 sid.inpe.br/sibgrapi/2018/10.18.12.41
%2 sid.inpe.br/sibgrapi/2018/10.18.12.41.19
%T Exploring Feature Distribution to Create Mid-level Representations: A Case Study in Human Action Recognition
%D 2018
%A Almeida, Raquel,
%A Jr. , Zenilton K. G. do Pratrocínio,
%A Guimarães, Silvio Jamil F.,
%@affiliation Pontifical Catholic University of Minas Gerais
%@affiliation Pontifical Catholic University of Minas Gerais
%@affiliation Pontifical Catholic University of Minas Gerais
%E Ross, Arun,
%E Gastal, Eduardo S. L.,
%E Jorge, Joaquim A.,
%E Queiroz, Ricardo L. de,
%E Minetto, Rodrigo,
%E Sarkar, Sudeep,
%E Papa, João Paulo,
%E Oliveira, Manuel M.,
%E Arbeláez, Pablo,
%E Mery, Domingo,
%E Oliveira, Maria Cristina Ferreira de,
%E Spina, Thiago Vallin,
%E Mendes, Caroline Mazetto,
%E Costa, Henrique Sérgio Gutierrez,
%E Mejail, Marta Estela,
%E Geus, Klaus de,
%E Scheer, Sergio,
%B Conference on Graphics, Patterns and Images, 31 (SIBGRAPI)
%C Foz do Iguaçu, PR, Brazil
%8 29 Oct.-1 Nov. 2018
%I Sociedade Brasileira de Computação
%J Porto Alegre
%S Proceedings
%K data representation, mid-level, bag-of-words, human action recognition.
%X Data representation is a critical task in many areas of computational studies, particularly in the case of visual data representation, in which subtleties can undermine the perception and interpretation of the visual content. In this study, it is proposed strategies to exploit visual mid-level representations, aiming to transform the detailed description extracted directly from the visual media into a simplified and discriminative representation. More specifically, the proposed strategies are delineated in Bag-of-Words mid-level representation model and are used to aggregate distribution information within partitions and regions of interest on feature space. Experiments on three well-known public datasets, namely, KTH, UCF Sports and UCF 11, demonstrated that feature points spatial distribution information is useful to create more discriminative representations. All three proposed representations were published and outperform, in terms of recognition rate, conventional strategies on BoW model and are, in many cases, superior or comparable with the state-of-the-art.
%@language en
%3 paperID5.pdf