- Notifications
You must be signed in to change notification settings - Fork0
Extending MaxViT (Multi-Axis Vision Transformer) to 3D Space
License
NotificationsYou must be signed in to change notification settings
nengwp/3D-MaxViT-pytorch
Folders and files
| Name | Name | Last commit message | Last commit date | |
|---|---|---|---|---|
Repository files navigation
Extending MaxViT (Multi-Axis Vision Transformer) to 3D Space
MaxViT:Multi-Axis Vision Transformer. In this paper author introduce an efficient and scalable attention model , call multi-axis attention, which consists of two aspects: blocked local and dilated global attention.
We extend MaxViT to 3D space to accommodate medical images such as CT and MRI. We implemented 3D Max-SA(3D Multi-axis self-attention) as shown in the figure to facilitate insertion into existing medical networks such as 3DUnet.
We refer to part of the code ofvit-pytorch/maxvit