A novel capsule network based on residual module and attention mechanism

Capsule network(CapsNet) is an improvement of the classic convolutional neural network(CNN), which has achieved good results in the field of image classification. However, the high computational complexity of the routing process hinders them from building larger models. Fortunately, reducing the complexity of CapsNet routing through spatial attention mechanisms has achieved significantly good results. To this end, we propose a capsule network based on residual pose routing and attention mechanism. While reducing the routing calculation complexity, the accuracy can be improved. Experimental results on MNIST, CIFAR-10 and SmallNORB show that the proposed method can achieve better classification results compared with traditional methods.