In this paper, we propose a conceptually simple but very effective attention module for Convolutional Neural Networks (ConvNets). In contrast to existing channel-wise and spatial-wise attention ...