Integrating Multiple Receptive Fields Through Grouped Active Convolution

Convolutional networks have achieved great success in various vision tasks. This is mainly due to a considerable amount of research on network structure. In this study, instead of focusing on architectures, we focused on the convolution unit itself. The existing convolution unit has a fixed shape and is limited to observing restricted receptive fields. In earlier work, we proposed the active convolution unit (ACU), which can freely define its shape and learn by itself. In this paper, we provide a detailed analysis of the previously proposed unit and show that it is an efficient representation of a sparse weight convolution. Furthermore, we extend an ACU to a grouped ACU, which can observe multiple receptive fields in one layer. We found that the performance of a naive grouped convolution is degraded by increasing the number of groups; however, the proposed unit retains the accuracy even though the number of parameters decreases. Based on this result, we suggest a depthwise ACU (DACU), and various experiments have shown that our unit is efficient and can replace the existing convolutions.

1 Introduction

Convolutional neural network (CNN) has become a major topic of deep learning, especially in visual recognition tasks. After the great success at the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) of 2012 [1], many efforts have been made to improve accuracy while reducing computational budgets by using CNN. The major focus for this research was on designing network architectures [2], [3], [4], [5], [6], [7]. Recently, attempts were made to automatically generate efficient network architectures [8], [9], and the generated networks achieved a better result than the conventional networks. This approach is yet very slow and difficult to train by using feasible amounts of resources but will affect the designing of networks. In such studies, components can be considered as more important factors than network construction.