Adversarial Residual Transformers For Multi-Modal Medical Imagesynthesis
Onat Dalmaz, Mahmut Yurt, Tolga Cukur
-
Members: FreeSPS
IEEE Members: $11.00
Non-members: $15.00Length: 00:04:12
28 Mar 2022
Medical image synthesis tasks are predominantly learned by Generative Adversarial Network (GAN) models based on convolutional backbones, which perform local processing with small filters. This inductive bias is sub-optimal for learning long-range spatial dependencies. To address this limitation, we propose a novel adversarial transformer model for medical image synthesis, ResViT, to synergistically combine convolution operators with residual transformer modules.
