As can be seen in the preceding screenshot, the shapes are described by their landmark sequences. One could use a program like GIMP or ImageJ as well as building a simple application in OpenCV, in order to annotate the training images. We will assume the user has completed this process and saved the points as sequences of x and y landmark positions for all training images in a text file, which will be used in our PCA analysis. We will then add two parameters to the first line of this file, which is the number of training images and the number of read columns. So, for k 2D points, this number will be 2*k.