Charla: Local Distribution of Descriptors on Concentric Hyperspheres in a Feature Space: A Case Study in Human Action Classification

Silvio Guimaraes, Profesor Pontifícia Universidade Católica de Minas Gerais
27 Julio, 2017 - 12:00
Auditorio Ramón Picarte, Piso 3, Edificio Norte
Prof. Benjamín Bustos


Silvio Jamil Ferzoli Guimaraes received his B.Sc. M.Sc. and D.Sc. degrees in Computer Science, from Universidade Federal de Vicosa (UFV), Universidade Estadual de Campinas (UNICAMP) and Universidade Federal de Minas Gerais (UFMG), in 1997, 1999 and 2003, respectively. Thanks to co-supervision program between UFMG and Université Marne-la-Vallée, he received another D.Sc. degree in Informatique. Silvio is a member of the Computer Science Department (DCC) staff, Pontifícia Universidade Católica de Minas Gerais (PUC Minas), Belo Horizonte-MG, Brazil (since 2002). He is the founder and header of the Audio-Visual Information Processing Laboratory. His research interests include digital image and video processing and computer vision applications, hierarchical information analysis, multimedia information systems, and content based information (image and video) retrieval.



Abstract:  In human action classification task, a video must be classified into a pre-determined class. To cope with this problem, we propose a mid-level representation which extends the Bag-of-Words formalism in order to better described the low-level features, exploring distance-to- codeword histograms. Our main contribution  is the assembly of low-level features by a mid-level representation enriched with information about distances between descriptors and codewords. The proposed representation takes into account volumes of hyper-regions obtained from hyperspheres centered at codewords.  Experimental results demonstrated that our strategy either improved or it is a competitive one in terms of classification rates for three well-known datasets.