Harmonic convolutional networks based on discrete cosine transform

作者：

Highlights：

• The harmonic block is designed to learn filter weights in the DCT domain.

• Harmonic CNNs are constructed by replacing the convolutional layers.

• Parameter learning in frequency domain improves performance.

• High-frequency parameter truncation can efficiently compress new or trained CNNs.

• The hamonic block can make a CNN invariant to illumination changes.

摘要

•The harmonic block is designed to learn filter weights in the DCT domain.•Harmonic CNNs are constructed by replacing the convolutional layers.•Parameter learning in frequency domain improves performance.•High-frequency parameter truncation can efficiently compress new or trained CNNs.•The hamonic block can make a CNN invariant to illumination changes.

论文关键词：Harmonic network,Convolutional neural network,Discrete cosine transform,Image classification,Object detection,Semantic segmentation

论文评审过程：Received 14 January 2021, Revised 16 December 2021, Accepted 7 April 2022, Available online 11 April 2022, Version of Record 21 April 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108707