In the last decade, convolutional neural networks (CNNs) have been the go-to architecture in computer vision, owing to their powerful capability in learning representations from images/videos.