A new method for automatically building statistical shape models from a set of training examples and in particular from a class of hands. In this method, landmark extraction is achieved using a self-organising neural network, the Growing Neural Gas (GNG), which is used to preserve the topology of any input space. Using GNG, the topological relations of a given set of deformable shapes can be learned. We describe how shape models can be built automatically by posing the correspondence problem on the behaviour of self-organising networks that are capable of adapting their topology to an input manifold, and due to their dynamic character to readapt it to the shape of the objects. Results are given for the
training set of hand outlines, showing that the proposed
method preserves accurate models.