Topic 05: Convolutional Neural Networks - Part I
This chapter introduce –convolutional neural networks (CNNs)– one of the most popular component of deep learning architecture. CNNs are widely applied in all type of domains such as natural language processing (NLP), audio, and time-series data. In this part, we introduce the CNNs, properties and components of CNN, differences between CNN and FCN as well as math behind the CNNs.
-
Chapter 05.01: Introduction of Convolution Neural Networks (CNNs)
This subchapter briefly covers the primary components of CNN architectures and explores applications of CNNs in fields like autonomous driving, medical imaging, and natural language processing.
-
Chapter 05.02: Convolutional Operation
This subchapter introduces convolutional operations, focusing on the role of filters in feature extraction. It explains how convolutional layers apply learned filters to images and also covers the 2D convolution operation in detail, illustrating the step-by-step computation process.
-
Chapter 05.03: Properties of Convolution
In this subchapter, we introduce sparse interactions, parameter sharing and the equivariance to translation, which are three important properties of convolution.
-
Chapter 05.04: CNN Components
This subchapter covers fundamental components of CNNs such as input channels, padding, stride, and pooling layers. Padding and stride settings control the dimensions and details retained in feature maps, while pooling layers i.a. reduce data size and improve computational efficiency. Together, these components enable CNNs to handle high-dimensional data effectively, capturing both local and global patterns in images.
-
Chapter 05.05: CNN Application
In this subchapter, we explore diverse applications of CNNs e.g. in visual recognition tasks, image classification, object detection, colorization, and semantic segmentation, illustrating their adaptability across domains.