FreConv: Frequency Branch-and-Integration Convolutional Networks
This work addresses a domain-specific problem for computer vision researchers by offering an incremental improvement to convolutional network design for better frequency utilization.
The paper tackles the problem of underutilizing frequency information in convolutional networks by proposing FreConv, a dual-branch module that extracts and integrates high- and low-frequency data, resulting in enhanced feature representation and reduced computational costs while consistently outperforming state-of-the-art baselines across various tasks.
Recent researches indicate that utilizing the frequency information of input data can enhance the performance of networks. However, the existing popular convolutional structure is not designed specifically for utilizing the frequency information contained in datasets. In this paper, we propose a novel and effective module, named FreConv (frequency branch-and-integration convolution), to replace the vanilla convolution. FreConv adopts a dual-branch architecture to extract and integrate high- and low-frequency information. In the high-frequency branch, a derivative-filter-like architecture is designed to extract the high-frequency information while a light extractor is employed in the low-frequency branch because the low-frequency information is usually redundant. FreConv is able to exploit the frequency information of input data in a more reasonable way to enhance feature representation ability and reduce the memory and computational cost significantly. Without any bells and whistles, experimental results on various tasks demonstrate that FreConv-equipped networks consistently outperform state-of-the-art baselines.