Fisher Vectors for Fine-Grained Visual Categorization

Jorge Sánchez 1 Florent Perronnin 2 Zeynep Akata 2, 3, *
* Corresponding author
3 LEAR - Learning and recognition in vision
Inria Grenoble - Rhône-Alpes, LJK - Laboratoire Jean Kuntzmann, INPG - Institut National Polytechnique de Grenoble
Abstract : The bag-of-visual-words (BOV) is certainly the most popular image representation to date and it has been shown to yield good results in various problems including Fine-Grained Visual Categorization (FGVC) [3, 4]. Our contribution is to show that the Fisher Vector (FV) - which describes an image by its deviation from an "average" model - is an excellent alternative to the BOV for the FGVC problem. In this extended abstract we first provide a brief introduction to the FV. We then present theoretical as well as practical motivations for using the FV for FGVC. We finally provide experimental results on four ImageNet subsets: fungus, ungulate, vehicle and ImageNet10K. Compared to [4] which uses spatial pyramid (SP) BOV representations, we report significantly higher classification accuracies. For instance, on ImageNet10K we report 16.7% vs 6.4% top-1 accuracy (a 160% relative improvement).
Document type :
Conference papers
Complete list of metadatas

https://hal.inria.fr/hal-00817681
Contributor : Thoth Team <>
Submitted on : Thursday, April 25, 2013 - 10:36:24 AM
Last modification on : Tuesday, February 12, 2019 - 10:30:05 AM
Long-term archiving on : Friday, July 26, 2013 - 4:02:34 AM

Files

Identifiers

  • HAL Id : hal-00817681, version 1

Collections

Citation

Jorge Sánchez, Florent Perronnin, Zeynep Akata. Fisher Vectors for Fine-Grained Visual Categorization. FGVC Workshop in IEEE Computer Vision and Pattern Recognition (CVPR), IEEE, Jun 2011, Colorado Springs, United States. ⟨hal-00817681⟩

Share

Metrics

Record views

757

Files downloads

756