Pooling and Strided Downsampling in Convolutional Networks

downsampling is what buys receptive field cheaply, and it is the half of the U-Net that the encoder still does

jeremy Video

Max Pooling in Convolutional Neural Networks to Reduce Layer Count

Max pooling is a downsampling mechanism in Convolutional Neural Networks (CNNs) designed to reduce feature map dimensions and mitigate issues such as diminishing gradients in deep architectures. This…