This is the Git repository for the program 'CNN assisted PSF localization (CAPL)' based on the paper Neural network-assisted localization of clustered point spread functions in single-molecule localization microscopy.
The program was developed in Python 3.8.8. Later versions of Python should be supported but have not been tested yet. If you are on a later version, give it a try!
I would suggest you use Anaconda / Miniconda and set up a virtual environment as:
conda create -n "myenv" python=3.8.8# replace "myenv" with your desired name.
After setting up the virtual environment, you can install the dependencies as:
-
pip install -r requirements.txt -
Additionally you will need to set up Fiji / ImageJ and install the ThunderSTORM plugin.
The program is divided into four parts:
- 01_training_data_generation.ipynb : this notebook generates the training files required to train the model.
- 02_training_model.ipynb : this notebook trains the CNN model
- 03_prediction.ipynb : this notebook is used to employ the trained model to predict super-resolved images from unseen data
- ImageVisualization.py : this script is reused from one of my old projects; used to create the super-resolved image from the detections
A step-by-step procedure for using each notebook is incorporated into the notebooks.
This code is prepared based on Deep-STORM and ZeroCostDL4Mic platform. Please also cite their work.
Thank you,
Pranjal Choudhury