GrabCut Technique

Before we go into the details of this amazing technique, we need to understand the term Image Segmentation, which will form the very basis of GrabCut. In layman's terms, an image segmentation task comprises dividing the pixels into different categories based on the class they belong to. Here, the classes we are looking for are just the background and foreground. This means that we want to segment the image by deciding what region of the image is part of the background and what region of the image is part of the foreground.

Note

In deep learning, image segmentation is a very big topic and is comprised of segmenting an image based on its classes, sometimes based on semantics, and so on. For now, we can consider GrabCut a simple background removal technique, though technically it can also be considered as an image segmentation technique with only two primary classes in mind – background and foreground. It's also important to note here that GrabCut...