|Topic:||Learning to segment from object counts|
|Supervisor:||Prof. Dr. Ing. Jan Kybic|
|Description:||The task is to learn to count objects in an image, when the only information available for the training images are the counts. What assumptions are needed to make it possible? We start from simpler cases, e.g. when the segmentation method has a small number of parameters or when pixel-wise annotations are available for at least some of the images and then work towards an efficient and practical method for more difficult cases.