Skip to content

Custom pytorch modules with sane default parameters useful for model prototyping. (U-Net, upsampling, downsampling...)

License

Notifications You must be signed in to change notification settings

vsitzmann/pytorch_prototyping

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

14 Commits
 
 
 
 
 
 

Repository files navigation

Pytorch modules with sane defaults for model prototyping

This is a number of pytorch modules (all based on prior work of the ML community) with sane default parameters that I find useful in model prototyping.

I'll continue to update this repository.

Contains:

  • 2d U-Net with different options for how the feature maps are upsampled (to prevent checkerboard artifacts.)
  • 3d U-Net
  • 2d downsampling network
  • 2d upsampling network with different options for how the feature maps are upsampled (to prevent checkerboard artifacts.)
  • 2d conv layer that pads to keep the spatial dimensions of the feature map constant, with reflection padding.

About

Custom pytorch modules with sane default parameters useful for model prototyping. (U-Net, upsampling, downsampling...)

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages