Semantic Segmentation
In Chapter 11, Object Detection, we discussed object detection as an important computer vision algorithm with diverse practical applications. In this chapter, we will discuss another related algorithm called Semantic Segmentation. If the goal of object detection is to perform simultaneous localization and identification of each object in the image, in semantic segmentation, the aim is to classify each pixel according to its object class.
Extending the analogy further, in object detection, we use bounding boxes to show results. In semantic segmentation, all pixels for the same object belong to the same category. Visually, all pixels of the same object will have the same color. For example, all pixels belonging to the soda can category will be blue in color. Pixels for non-soda can objects will have a different color.
Similar to object detection, semantic segmentation has many practical applications. In medical imaging, it can be...
