GPU-Net: Lightweight U-Net with more diverse features

TL;DR

GPU-Net is a lightweight U-Net variant that incorporates Ghost modules and ASPP to learn more diverse features, achieving superior performance with significantly fewer parameters and FLOPs, and can enhance existing segmentation models.

Contribution

Introduces GPU-Net with Ghost modules and ASPP, offering a more efficient and effective segmentation network with broader feature learning capabilities.

Findings

Over 4x fewer parameters than traditional U-Net

2x reduction in FLOPs while maintaining performance

Plug-and-play module improves existing segmentation methods

Abstract

Image segmentation is an important task in the medical image field and many convolutional neural networks (CNNs) based methods have been proposed, among which U-Net and its variants show promising performance. In this paper, we propose GP-module and GPU-Net based on U-Net, which can learn more diverse features by introducing Ghost module and atrous spatial pyramid pooling (ASPP). Our method achieves better performance with more than 4 times fewer parameters and 2 times fewer FLOPs, which provides a new potential direction for future research. Our plug-and-play module can also be applied to existing segmentation methods to further improve their performance.