Product Images Classification with Mul-tiple Features Combination

Shi-jie Jia, Xiang-wei Kong, Hai-yan Fu, Guang Jin
2010 Proceedings of the 2010 International Conference on E-Business Intelligence   unpublished
Visual-based automatic product image classification is a great need and challenge work for e-commerce field. Previous work tested few number product categories with one or two descriptors. For the task of product classification over large number categories, we employed kernel-based SVM classifier combining multiple features, including one global descriptor and three complimentary local descriptors. Furthermore, we investigate four ways to combine discriminative features for SVM classifier.
more » ... iments on the product image dataset (PI 100) showed the performance improved significantly by features fusion.
doi:10.2991/icebi.2010.66 fatcat:xgzuhislpbakpgqoe6vkt6amsu