Research on U-Net Improvement by Attention and Residual Block

Abstract:
In recent years, research in the area of the medical image process has picked up a fervent pace, especially after U-Net has been put forward. U-Net’s ingenious architecture and ideas have been widely cited by people, and various neural networks have been proposed, such as Res- UNet, Attention UNet, 3D U-Net, just to name a few. It is indeed evident that people have derived various network architectures based on U-Net. It is indeed evident that there is various neural networks derived from U-Net. Hence, in order to avoid repeating the past work and give a review of recent achievements, an overview is necessary. Thus, this paper will summarize the outstanding work that people have used the attention mechanism and residual block to improve U-Net in the past.