Image Classification with the Fisher Vector: Theory and Practice

• Autores: Sanchez, Jorge Adrian; Perronnin, Florent; Mensink, Thomas; Verbeek, Jakob
• Año de publicación: 2013
• Idioma: inglés
• Tipo de recurso: artículo
• Materia: Image Classification; Large-Scale Classification; Bag-Of-Visual Words; Fisher Vector
• Institución: Consejo Nacional de Investigaciones Científicas y Técnicas

A standard approach to describe an image for classification and retrieval purposes is to extract a set of local patch descriptors, encode them into a high dimensional vector and pool them into an image-level signature. The most common patch encoding strategy consists in quantizing the local descriptors into a finite set of prototypical elements. This leads to the popular Bag-of-Visual words representation. In this work, we propose to use the Fisher Kernel framework as an alternative patch encoding strategy: we describe patches by their deviation from an “universal” generative Gaussian mixture model. This representation, which we call Fisher vector has many advantages: it is efficient to compute, it leads to excellent results even with efficient linear classifiers, and it can be compressed with a minimal loss of accuracy using product quantization. We report experimental results on five standard datasets—PASCAL VOC 2007, Caltech 256, SUN 397, ILSVRC 2010 and ImageNet10K— with up to 9M images and 10K classes, showing that the FV framework is a state-of-the-art patch encoding technique.
Fil: Sanchez, Jorge Adrian. Consejo Nacional de Investigaciones Científicas y Técnicas. Centro Científico Tecnológico Córdoba. Centro de Investigación y Estudios de Matemática de Córdoba(p); Argentina
Fil: Perronnin, Florent . Xerox Research Centre Europe; Francia
Fil: Mensink, Thomas. University of Amsterdam. Inteligent Systems Lab Amsterdam; Países Bajos
Fil: Verbeek, Jakob. LEAR Team, INRIA Grenoble; Francia