Which of the following is not a valid use of Generative Adversarial Networks? a. To generate a random song after being trained from a large collection of songs. b. To classify an image after being trained on a large collection of labelled images. C. To generate an X-ray image of the chest based upon a text description of the lung disease the person is suffering. d. Given one half of an image generate the second half. Suppose you are asked to find the center position of every person in a soccer field from a given image. Which of the following models would be the most useful for this problem? a. A model trained for classification, where person is one of the pre-trained classes. b. A model trained for semantic segmentation, where person is one of the pre-trained classes. C. A model trained for object detection, where person is one of the pre-trained classes. A model trained for regression, where the output is the number of people in the image.