2. This part explores the use of model-based clustering for image segmentation. The data is a 400 x 300 RGB image of a person serving in table tennis, and is found in tt. rda. The pixel locations are given by tt$x and ttsy and the observed variables are tt$R, tt$G and tt$8. You will not need to scale the variables before clustering. For reference, the actual image is tt. png. (a) Carry out model-based clustering, using BIC for model selection. This may take some time! Give a scatterplot (in terms of pixel locations) that is colour coded by cluster label. (Don't worry if it is upside-down.) What explains the clusters you observe? [7 marks] (b) Carry out model-based clustering, using ICL for model selection. Give a scatterplot that is colour coded by cluster label (you may need to explicitly fit the selected model first). What differences do you observe with the first clustering? [8 marks] Total: 30 marks