Dilated Convolution in Deep Learning

Dilated convolution is the concept of skipping the pixel while convoluting the kernel with the input.

It is also known as the pixel-skipping algorithm. the factor by which pixels are skipped is known as Dilation Factor (l)

Features of Dilated Convolution

• It helps to cover a larger area on the image with less computation
• It down-samples the features without pooling

Articles to Read

1. https://www.geeksforgeeks.org/dilated-convolution/
2. https://towardsdatascience.com/review-dilated-convolution-semantic-segmentation-9d5a5bd768f5

Standard Convolution vs Dilated Convolution