Interleaved Group Convolutions | RDL Research Database

Abstract

In this paper, we present a simple and modularized neural network architecture, named interleaved group convolutional neural networks (IGCNets). The main point lies in a novel building block, a pair of two successive interleaved group convolutions: primary group convolution and secondary group convolution. The two group convolutions are complementary: (i) the convolution on each partition in primary group convolution is a spatial convolution, while on each partition in secondary group convolution, the convolution is a point-wise convolution; (ii) the channels in the same secondary partition come from different primary partitions. We discuss one representative advantage: Wider than a regular convolution with the number of parameters and the computation complexity preserved. We also show that regular convolutions, group convolution with summation fusion, and the Xception block are special cases of interleaved group convolutions. Empirical results over standard benchmarks, CIFAR-10, CIFAR-100, SVHN and ImageNet demonstrate that our networks are more efficient in using parameters and computation complexity with similar or higher accuracy.

Keywords

Convolution (computer science)Partition (number theory)ComputationConvolution powerConvolutional neural networkGroup (periodic table)Block (permutation group theory)Computer scienceConvolution theoremMathematicsOverlap–add methodAlgorithmArtificial neural networkArtificial intelligenceCombinatoricsFourier transformMathematical analysis

Affiliated Institutions

Related Publications

Xception: Deep Learning with Depthwise Separable Convolutions

François Chollet

We present an interpretation of Inception modules in convolutional neural networks as being an intermediate step in-between regular convolution and the depthwise separable convo...

2017 17644 citations

Xception: Deep Learning with Depthwise Separable Convolutions

François Chollet

We present an interpretation of Inception modules in convolutional neural networks as being an intermediate step in-between regular convolution and the depthwise separable convo...

2016 arXiv (Cornell University) 357 citations

clcNet: Improving the Efficiency of Convolutional Neural Network Using Channel Local Convolutions

Dongqing Zhang

Depthwise convolution and grouped convolution has been successfully applied to improve the efficiency of convolutional neural network (CNN). We suggest that these models can be ...

2018 15 citations

Densely Connected Convolutional Networks

Gao Huang , Zhuang Liu , Laurens van der Maaten +1 more

Recent work has shown that convolutional networks can be substantially deeper, more accurate, and efficient to train if they contain shorter connections between layers close to ...

2017 42167 citations

Network In Network

Min Lin , Qiang Chen , Shuicheng Yan

Abstract: We propose a novel deep network structure called In Network (NIN) to enhance model discriminability for local patches within the receptive field. The conventional con...

2014 arXiv (Cornell University) 1037 citations

Publication Info

Year: 2017
Type: article
Citations: 327
Access: Closed

External Links

View on DOI.org

Social Impact

Altmetric

Interleaved Group Convolutions

PlumX Metrics

Social media, news, blog, policy document mentions

Citation Metrics

327

OpenAlex

Cite This

APA Style

                            
                                    Ting Zhang, 
                                
                                    Guo-Jun Qi, 
                                
                                    Bin Xiao
                                
                                et al.
                            
                            (2017). 
                            Interleaved Group Convolutions. 
                            
                            .
                            https://doi.org/10.1109/iccv.2017.469

Identifiers

DOI: 10.1109/iccv.2017.469