How to localize an object in an image using convolutional neural networks?

Question

Neural networks

Started learning convolutional neural networks. I figured out how convolutional, pooling and relu layers work, I imagine how to train the network to determine whether there is an object in the image or not. But how to use the data from these layers to locate the object in the photo? On the example of the Helen base , there are already marked points for the contour of the face, nose, mouth and eyes. What should be the structure and principle of operation of the last layer? Thanks in advance