Object class recognition is a fundermental challenge to a computer vision system. In this presentation, I will present a bag of local features model for object class recognition from natural still images. The basic idea of this model is treating an image as an unordered set of local features. After quantitize the local features, a histogram over the visual vocabulary is used as the features of images. These features, combining with SVM, give out very good performance on some challenging datasets.

Return to Conference Programme