2. Semantic segmentation network

From the previous section, we learned that the semantic segmentation network is a pixel-wise classifier. The network block diagram is shown in Figure 12.2.1. However, unlike a simple classifier (for example, the MNIST classifier in Chapter 1, Introducing Advanced Deep Learning with Keras and Chapter 2, Deep Neural Networks), where there is only one classifier generating a one-hot vector as output, in semantic segmentation, we have parallel classifiers running simultaneously. Each one is generating its own one-hot vector prediction. The number of classifiers is equal to the number of pixels in the input image or the product of image width and height. The dimension of each one-hot vector prediction is equal to the number of stuff object categories of interest.

Figure 12.2.1: The semantic segmentation network can be viewed as a pixel-wise classifier. Best viewed in color. The original images can be found at https://github.com/PacktPublishing...