Title : Manifold learning and Random Projections for multi - view object recognition

Recognizing objects from different viewpoints is a challenging task. One approach for handling this task is to model the appearance of an object under different viewing conditions using a low dimensional subspace. Manifold learning describes the process by which this low dimensional embedding can be generated. However, manifold learning is an unsupervised method and thus gives poor results on classification tasks. To address this impediment of manifold learning, we investigated the combination of manifold learning and distance metric learning for the generation of a representation that is both discriminative and informative. In the proposed system, initial dimensionality reduction is achieved using random projections, a computationally efficient and data independent linear transformation. Distance metric learning is then applied to increase the separation between classes and improve the accuracy of nearest neighbor classification. Finally, a manifold learning method is used to generate a mapping between the randomly projected data and the low dimensional manifold. We demonstrate that this approach is effective for multi view face recognition. Furthermore, we show that random projections can be applied as an initial step without significantly affecting the classification accuracy. Although we provide results regarding the classification of human faces from different viewpoints, the same approach can be used for the classification of generic objects under different viewing conditions.