qnano/drift-estimation

Drift estimation

Example code for "Drift correction in localization microscopy using entropy minimization"

Article link: https://www.biorxiv.org/content/10.1101/2021.03.30.437682v1

Google colab example (For some reason the CPU multithreading runs very slow on linux though) https://colab.research.google.com/github/qnano/drift-estimation/blob/master/dme_example_colab.ipynb

Build from source (tested on windows and linux)

DISCLAIMER: My cmake file was created entirely by trial and error, please let me know if I'm doing something weird.

• On Windows, install cmake and make sure it is available in the path. Then, run build_windows.bat
• On Linux (tested on Ubuntu 20.04):
  • Install CUDA from the nVidia website. Install cmake and use "cmake .", followed by "make" in the repository root.
  • Currently, CPU-only build is not compiling, you need CUDA.
  • This should build ./bin/release/libdme_cuda.so
  • If anything goes wrong, cmake intermediary files have to be removed before starting over (CMakeFiles directory and CMakeCache.txt)
  • dme_example.py should now run without errors.
• On MacOS, CUDA is no longer supported by nVidia, so only the CPU version can be used.
  • Thanks to Duncan Ryan for help getting the GCC build working and confirming it runs on MacOS!

To run dme_example.py using pre-build Windows binaries:

1. Install python. Anaconda is recommended: https://www.anaconda.com/distribution/
2. Create a virtual environment, such as an anaconda environment:

conda create -n myenv anaconda python=3.8
conda activate myenv

3. Install required pip packages, installing these within the virtual environment:

pip install tqdm scipy numpy matplotlib 

4. Run the example code:

python dme_example.py