Skip to content

mmiv-center/fake-cells

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Generate Fake Cells

For deep learning (pre-training) we want to have some pristine data with a known ground truth. The model should be adjusted later to real data but might need less of it and/or train faster.

2000 example training images

The above image is a montage of 20 volumes with 100 slices each. Every pair of rows corresponds to one volume stack. The circular vignette has only been added to this montage, it is not part of the training data.

Our model cells have

  • cell walls
  • a tiny nucleus

We start backwards, instead of segmenting an image we create a segmentation and distort it to look like an image. In that way the ground truth image is the segmentation and the generated image is a representation based on some simple imaging physics like a 3d point-spread function and some photon-shot noise. Our generated images look (remotely) like light-microscopy images.

About

Cell separation demo using de-novo data

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published