Exploiting Cyclic Symmetry in Convolutional Neural Networks

2016/02/08 by Sander Dieleman, Dieleman, Sander, Jeffrey De Fauw +3 · 1 voice · 22 citations
Computer Science · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Image Processing and 3D Reconstruction #Image and Object Detection Techniques #Machine Learning (cs.LG) #Medical Image Segmentation Techniques #Neural and Evolutionary Computing (cs.NE) #cs.CV #cs.LG #cs.NE

paper · pdf · doi:10.48550/arxiv.1602.02660

10 pages, 6 figures, accepted for publication at ICML 2016

openalex publication_date 2016/02/08 · arxiv published 2016/02/08 · arxiv created 2016/05/26 · arxiv updated 2016/05/26 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Many classes of images exhibit rotational symmetry. Convolutional neural networks are sometimes trained using data augmentation to exploit this, but they are still required to learn the rotation equivariance properties from the data. Encoding these properties into the network architecture, as we are already used to doing for translation equivariance by using convolutional layers, could result in a more efficient use of the parameter budget by relieving the model from learning them. We introduce four operations which can be inserted into neural network models as layers, and which can be combined to make these models partially equivariant to rotations. They also enable parameter sharing across different orientations. We evaluate the effect of these architectural modifications on three datasets which exhibit rotational symmetry and demonstrate improved performance with smaller models.

Citations

• Very Deep Convolutional Networks for Large-Scale Image Recognition

Cited by

• Accelerated Rotation-Invariant Convolution for UAV Image Segmentation
• Can we have it all? On the Trade-off between Spatial and Adversarial Robustness of Neural Networks
• NeRV: Neural Representations for Videos
• Incorporating Symmetry into Deep Dynamics Models for Improved Generalization
• Spin-Weighted Spherical CNNs
• Automatic Analysis System of Calcaneus Radiograph: Rotation-Invariant Landmark Detection for Calcaneal Angle Measurement, Fracture Identification and Fracture Region Segmentation
• Learning Equivariant Representations
• B-Spline CNNs on Lie Groups
• Learnable Gabor modulated complex-valued networks for orientation robustness
• InterAct: Advancing Large-Scale Versatile 3D Human-Object Interaction Generation
• Equivariant Transformer Networks
• Hierarchical Adaptive networks with Task vectors for Test-Time Adaptation
• The Convolution Exponential and Generalized Sylvester Flows
• Learning Equivariant Energy Based Models with Equivariant Stein Variational Gradient Descent
• Equivariance Through Parameter-Sharing
• A Data and Compute Efficient Design for Limited-Resources Deep Learning
• Transformation Consistent Self-ensembling Model for Semi-supervised Medical Image Segmentation
• Doubly Convolutional Neural Networks
• Trajectory Prediction using Equivariant Continuous Convolution
• Retinal vessel segmentation based on Fully Convolutional Neural Networks
• Symmetry reduction for deep reinforcement learning active control of chaotic spatiotemporal dynamics
• Learning Compressed Transforms with Low Displacement Rank

Discussions

• Nice list! Couldn't help but notice our concurrent work with Cohen et al. 2016 is missing though 😁 arxiv.org/abs/1602.02660 [bsky, 0 points, 0 comments]

Related

• Rotation Equivariance and Invariance in Convolutional Neural Networks
• Deep Rotation Equivariant Network
• Rotation Invariance Neural Network
• Harmonic Networks: Deep Translation and Rotation Equivariance
• Rotated Ring, Radial and Depth Wise Separable Radial Convolutions