Learning and caricaturing the face space using self-organization and Hebbian learning for face processing

This paper shows a self-organized system designed to obtain compressed representations of instances of a population of visual forms. It is shown how, when applied to face shape information, the system evolves into a prototype of the population and induces automatic warping, or caricaturing, transformations where geometrical differences between forms are increased, improving, as a consequence, recognition performance. In this way, the proposed system provides a unified account for the whole chain of face processing tasks including data compression, detection, and recognition. Experimental data is presented to show the feasibility of our approach in terms of performance and robustness to changes in illumination and face expressions.