I summarized from some other web pages: Pros: Deep network because the output dimension could be constant after convolution [Saving image border information](https://stats.stackexchange.com/questions/246512/convolutional-layers-to-pad-or-not-to-pad) Cons: [Waste of computational resources](https://www.quora.com/Which-effect-does-sequence-padding-have-on-the-training-of-a-neural-network)