Skip to content

Godofnothing/Segmentation_and_StyleTransfer

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Segmentation and StyleTransfer


Implementation of localized style transfer from the paper Localized style transfer.

Here the simplified approach is adopted - we do not use the CRF (conditional random fields), which are used for smoothing and general improvement of the segmentation task.

In order to make a style transfer one needs to pass a content image, obtain a mask on it via evaluation on the pretrained segmentation network, and the style image.

The current implementation uses MaskRCNN with resnet-50 backbone and vgg19 backbone for the style transfer.

Function from detection directory are copied from [PyTorch vision] https://github.com/pytorch/vision

About

Localized style transfer on segmentated image

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published