Integrating Multiple Receptive Fields through Grouped Active Convolution

TL;DR

This paper introduces a grouped active convolution unit (ACU) that can observe multiple receptive fields simultaneously, improving efficiency and accuracy in convolutional networks without increasing parameters.

Contribution

It extends the ACU to a grouped version, enabling multiple receptive fields in one layer, and proposes a depthwise ACU that maintains performance with fewer parameters.

Findings

Grouped ACU retains accuracy with fewer parameters.

Naive grouped convolution performance degrades with more groups.

Depthwise ACU is efficient and effective as a replacement for existing convolutions.

Abstract

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, and various experiments have shown that our unit is efficient and can replace the existing convolutions.