Abstract

In this thesis we address the problem of image and video understanding and specifically, we tackle the problem with machine learning techniques. The primary techniques harnessed in our work are comprised of feature selection, semi-supervised learning, intermediate representation learning and knowledge adaptation. The final result of this thesis delivers a comprehension of how we can improve multimedia analysis through a variety of machine learning techniques. From the representation perspective, feature selection is potentially helpful. From the classification perspective, semi-supervised learning and transfer learning both bring in reasonable performance by using only few labeled training data.