Visual categorization with bags of keypoints

Authors: Chris Dance, Jutta Willamowski, Lixin Fan, Cedric Bray, Gabriela Csurka
ECCV International Workshop on Statistical Learning in Computer Vision, Prague, 2004.
We present a method for generic visual categorization. This technique exploits an analogy with learning methods for text categorization based on the simple bag of words approach. Two key novel aspects of this approach are that it handles multiple image categories simultaneously and that it is intrinsically invariant to affine image transformations. Results are presented for simultaneously classifying seven semantic visual categories using Naive Bayes techniques.
Year: 2004
Report number: 2004/010


2004_010.pdf (798.13 kB)