In this article, Marcel Luethi and Ghazi Bouabene look back at the concepts that they covered this week.
In this week we have taken a look at the main concepts and basic notions of shape modelling.
We have discussed what we mean by the term shape and how shapes can be represented using a set of boundary points. We have seen that we can characterize a shape family by defining a probability distribution, which assigns to every shape (i.e. to every configuration of the boundary points) a probability. Shape instances with a higher probability are more likely to belong to the shape family than unlikely instances.
We have already derived a first shape model, where we have used only two measurements (the length and the span of the hand) to represent the shape. Although this model is overly simplistic, it illustrates many of the ideas and concepts that we will develop in this course. The main difference to the models that we will deal with in the rest of this course is, that we will not only work with two measurements, but with all the points that represent the shape. For this we will need to discuss how we can represent the variations in the boundary points and how we can formulate a probabilistic model of these variations. This will be the main topic of next week.
By now you should also have ScalismoLab running and you may even have experimented a bit with it already. In the next few weeks, ScalismoLab will play an important role, as it will allow you to explore the theory, to visualize the otherwise abstract concepts, and to experiment with your own shape models.