Skip to content

Tools to create patches and build CSBDeep deep learning model for md-SR imaging

License

Notifications You must be signed in to change notification settings

nriss/generate-imaging-dataset

Repository files navigation

DATASET MICROSCOPY

Python script used to generate patches and create CSBDeep model applied to super resolution microscopy.

  1. Create patches where common spots have been found between image stack pairs.
  2. Build CSBDeep model
  3. Test CSBDeep model on images

Getting started

To work with DatasetMicroscopy :

  1. Save your image files (.tif) in a data directory
  • target for great quality images
  • source for poor quality images

Example of directory

  • data/target/{file1.ome.tif, file2.ome.tif, file3.ome.tif, file4.ome.tif}
  • data/source/{file1.ome.tif, file2.ome.tif, file3.ome.tif, file4.ome.tif}

The equivalent image stacks must have the same names between target and source directories

  1. Run generateData.py (python generateData.py)

This script will create a npz file containing pair of image patch needed to build the model

  1. Run createModel.py (python createModel.py)

The script will call CSBDeep functions and use the npz file to train the CARE model (cf CSBDeep getting started in https://github.com/CSBDeep/CSBDeep).

  1. Run constructImage.py (python constructImage.py)

This function allow to reconstruct an entire stack of images from the CSBDeep model previously trained

About

This tool uses modified code of picasso (https://github.com/jungmannlab/picasso) for spot localization and modified patches creation function from CSBDeep (https://github.com/CSBDeep/CSBDeep).

Tools developed by Nicolas Riss under the supervison of Julien Godet.

About

Tools to create patches and build CSBDeep deep learning model for md-SR imaging

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages