Convolutional neural networks (CNNs) utilize convolutional operations to extract useful information from data that has a topology associated with it. This works best for image and audio data. The input image, when passed through a convolution layer, produces several output images, known as output feature maps. The output feature maps detect features. The output feature maps in the initial convolutional layer may learn to detect basic features, such as edges and color composition variation.
The second convolutional layer may detect slightly more complicated features, such as squares, circles, and other geometrical structures. As we progress through the neural network, the convolutional layers learn to detect more ...
