Portrait Segmentation Challenge

The course project will be portrait segmentation. Portrait segmentation refers

to the separation of the portrait subject and background in an image $($see

Figure $1).$ Your algorithms will be evaluated on unseen portrait samples using

precision and recall and F$-$measures and run times of your method. You will

compete with each other in getting a better segmentation algorithm for portrait

images. You can use the training portrait images where you can evaluate your

performance yourself, but final evaluation will be on another set of portrait

images. You must use Matlab to implement your codes. To segment leaves

from background, you can use any image segmentation algorithm from the

literature. While choosing the algorithm, you should consider the problems in

portrait images.

The project zip folder contains $2$ subfolders with rgb$\_$images, and

segmented$\_$images. The rgb$\_$images folder contains RGB portrait images

and the segmented$\_$images folder contains segmented version of the original

portrait images in the rgb$\_$image folder.