Image Categorization with Semi-Supervised Learning
Zhenghua Yu · 2006
This paper addresses the problem of categorizing/classifying images, with an emphasis on utilizing unlabeled image data to achieve higher classification accuracy. The main contribution of this paper is two-fold: firstly we introduce graph based semi-supervised learning to the problem of image categorization. Secondly we propose a novel neighborhood preserving graph-based semi-supervised learning method. Experiments of applying the proposed method to categorize image data demonstrated its effectiveness.