Given an appearance model, this model can be further extended and be used for image interpretation. One can take the model and select a set of parameter values which make the model resemble a given image. It is a case of matching to an image for analysis purposes and it is treated as an optimisation problem. This process is sometimes known and referred to as ``model fitting'' because a model gets manipulated until it best fits an image. Matching (or fitting) is made possible thanks to a process characterised by model training. It is the first step which involves extending an appearance model and having it equipped with additional knowledge (learning of how to fit the model). The second step involves searching for matches in unseen images.