Instance segmentation using the U-net architecture

So far, in the previous two chapters, we have learned about detecting objects and also about identifying the bounding boxes within which the objects within an image are located. In this section, we will learn about performing instance segmentation, where all the pixels belonging to a certain object are highlighted while every other pixel isn't (this is similar to masking all the other pixels that do not belong to an object with zeros and masking the pixels that belong to the object with pixel values of one).

Getting ready

To perform instance segmentation, we will perform the following:

1. Work on a dataset that has the input image and the corresponding masked image of the...