Convolutional Neural Networks and Architectures

This chapter briefly introduces Convolutional Neural Networks (CNNs). One of the first CNNs is proposed in [41] (known as LeNet) to deal with handwriting recognition task. After that, CNN becomes the most popular deep neural network model to process visual data, including images and videos. Until now, a lot of modern fundamental computer vision systems, e.g., image classification and face recognition, are usually built upon convolutional networks.