Advanced Convolutional Networks

In Chapter 2, Understanding Convolutional Networks, we discussed the building blocks of convolutional neural networks (CNNs) and some of their properties. In this chapter, we'll go a step further and talk about some of the most popular CNN architectures. These networks usually combine multiple primitive convolution and/or pooling operations in a novel building block that serves as a base for a complex architecture. This allows us to build very deep (and sometimes wide) networks with high representational power that perform well on complex tasks such as ImageNet classification, image segmentation, speech recognition, and so on. Many of these models were first released as participants in the ImageNet challenge, which they usually won. To simplify our task, we'll discuss all architecture within the context of image classification. We&apos...