Deep Rotation Equivariant Network
arXiv:1705.08623
Abstract
Recently, learning equivariant representations has attracted considerable research attention. Dieleman et al. introduce four operations which can be inserted into convolutional neural network to learn deep representations equivariant to rotation. However, feature maps should be copied and rotated four times in each layer in their approach, which causes much running time and memory overhead. In order to address this problem, we propose Deep Rotation Equivariant Network consisting of cycle layers, isotonic layers and decycle layers. Our proposed layers apply rotation transformation on filters rather than feature maps, achieving a speed up of more than 2 times with even less memory overhead. We evaluate DRENs on Rotated MNIST and CIFAR-10 datasets and demonstrate that it can improve the performance of state-of-the-art architectures.
References in corpus (7)
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- Group Equivariant Convolutional Networks
- Rotation equivariant vector field networks
- Harmonic Networks: Deep Translation and Rotation Equivariance
- Learning to Extract Motion from Videos in Convolutional Neural Networks
- Flip-Rotate-Pooling Convolution and Split Dropout on Convolution Neural Networks for Image Classification
- TI-POOLING: transformation-invariant pooling for feature learning in Convolutional Neural Networks