18
28

FsaNet: Frequency Self-attention for Semantic Segmentation

Abstract

Considering the spectral properties of images, we propose a new self-attention mechanism with highly reduced computational complexity, up to a linear rate. To better preserve edges while promoting similarity within objects, we propose individualized processes over different frequency bands. In particular, we study a case where the process is merely over low-frequency components. By ablation study, we show that low frequency self-attention can achieve very close or better performance relative to full frequency even without retraining the network. Accordingly, we design and embed novel plug-and-play modules to the head of a CNN network that we refer to as FsaNet. The frequency self-attention 1) requires only a few low frequency coefficients as input, 2) can be mathematically equivalent to spatial domain self-attention with linear structures, 3) simplifies token mapping (1×11\times1 convolution) stage and token mixing stage simultaneously. We show that frequency self-attention requires 87.29%90.04%87.29\% \sim 90.04\% less memory, 96.13%98.07%96.13\% \sim 98.07\% less FLOPs, and 97.56%98.18%97.56\% \sim 98.18\% in run time than the regular self-attention. Compared to other ResNet101-based self-attention networks, \ourM achieves a new \sArt result (83.0%83.0\% mIoU) on Cityscape test dataset and competitive results on ADE20k and VOCaug. \ourM can also enhance MASK R-CNN for instance segmentation on COCO. In addition, utilizing the proposed module, Segformer can be boosted on a series of models with different scales, and Segformer-B5 can be improved even without retraining. Code is accessible at \url{https://github.com/zfy-csu/FsaNet

View on arXiv
Comments on this paper