Object categorization by learned universal visual dictionary

DMCA

Object categorization by learned universal visual dictionary (2005)

Cached

Download Links

Other Repositories/Bibliography

DBLP

by J. Winn , A. Criminisi , T. Minka

Venue:	IN ICCV
Citations:	301 - 8 self

BibTeX

@INPROCEEDINGS{Winn05objectcategorization,
    author = {J. Winn and A. Criminisi and T. Minka},
    title = {Object categorization by learned universal visual dictionary},
    booktitle = {IN ICCV},
    year = {2005},
    pages = {1800--1807},
    publisher = {IEEE Computer Society}
}

OpenURL

Abstract

This paper presents a new algorithm for the automatic recognition of object classes from images (categorization). Compact and yet discriminative appearance-based object class models are automatically learned from a set of training images. The method is simple and extremely fast, making it suitable for many applications such as semantic image retrieval, web search, and interactive image editing. It classifies a region according to the proportions of different visual words (clusters in feature space). The specific visual words and the typical proportions in each object are learned from a segmented training set. The main contribution of this paper is two fold: i) an optimally compact visual dictionary is learned by pair-wise merging of visual words from an initially large dictionary. The final visual words are described by GMMs. ii) A novel statistical measure of discrimination is proposed which is optimized by each merge operation. High classification accuracy is demonstrated for nine object classes on photographs of real objects viewed under general lighting conditions, poses and viewpoints. The set of test images used for validation comprise: i) photographs acquired by us, ii) images from the web and iii) images from the recently released Pascal dataset. The proposed algorithm performs well on both texture-rich objects (e.g. grass, sky, trees) and structure-rich ones (e.g. cars, bikes, planes).

Keyphrases

learned universal visual dictionary object categorization test image object class new algorithm user selects segmented training feature space web search discriminative appearance-based object class model object instance merge operation large difference web site pair-wise merging novel statistical measure training image final visual word general lighting condition pascal dataset object class label compact visual dictionary many application different visual word interactive demo validation comprise visual appearance main contribution semantic image retrieval algorithm performs real object typical proportion interactive object categorization demo application structure-rich one texture-rich object specific visual word exemplar snapshot visual word high classification accuracy training set interactive image automatic recognition correct class label

DMCA