Name
Abstract | ||
|---|---|---|
We present a novel image classification system that is built around a core of trainable filter ensemble that we call Volterra kernel classifiers. Our system treats images as a collection of possibly overlapping patches and is composed of three components: (a) A scheme for single patch classification that seeks a smooth, possibly non-linear, functional mapping of the patches into a range space, where patches of the same class are close to one another, while patches from different classes are far apart -- in the $L_2$ sense. This mapping is accomplished using trainable convolution filters (or Volterra kernels) where the convolution kernel can be of any shape or order; (b) Given a corpus of Volterra classifiers with various kernel orders and shapes for each patch, a boosting scheme for automatically selecting the best weighted combination of the classifiers to achieve higher per-patch classification rate; (c) A scheme for aggregating classification information obtained for each patch via voting for the parent image classification. We demonstrate effectiveness of our method using face recognition and provide extensive experiments on five benchmark face datasets. Our technique, which falls into the class of embedding-based face image discrimination methods, consistently outperforms various state-of-the-art methods in same category. |
Year | DOI | Venue |
|---|---|---|
2012 | 10.1109/TPAMI.2011.225 | IEEE Trans. Pattern Anal. Mach. Intell. |
Keywords | DocType | Volume |
face recognition,higher per-patch classification rate,parent image classification,novel image classification system,single patch classification,volterra kernel,volterra classifier,classification information,trainable convolution filters,convolution kernel,volterra kernel classifier,merl dome benchmark face,kernel,feature extraction,image classification,shape,convolution,boosting,embedded systems | Journal | 34 |
Issue | ISSN | Citations |
7 | 1939-3539 | 15 |
PageRank | References | Authors |
0.55 | 26 | 4 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Ritwik Kumar | 1 | 86 | 8.23 |
| Arunava Banerjee | 2 | 313 | 29.18 |
| B.C. Vemuri | 3 | 4208 | 536.42 |
| Hanspeter Pfister | 4 | 5933 | 340.59 |