Star reduction in deep sky images

⚠️	This repository has been archived. Please checkout my new shiny tool (Starrem2k13) for star removal : https://code2k13.github.io/starrem2k13/

A GAN model trained to remove stars from astronomical images. Code was inspired from a sample at Tensorflow's website. The training data consists of only two images. One image of the Helix nebula and another is a starmap that was created from a star cluster image. Here is how the results look like:

Running using Docker (Recommended)

docker run   -v $PWD:/usr/src/app/starreduction/data  \
-it code2k13/starreduction \
/bin/bash -c "./removestars.py ./data/example.jpg  ./data/example_starless.jpg"

Note that $PWD refers to your current working directory. In the above example it is assumed that the file example.jpg resides in your current working directory. This directory is mounted as a volume with the path /usr/src/app/starreduction/data inside the docker container. The output image example_starless.jpg will also be written to same directory.

Running without Docker

Clone the repository and navigate to the 'starreduction' folder. Install required packages :

pip install -r requirements.txt

Additionally you may also have to install lfs support for git

git clone https://github.com/code2k13/starreduction.git
cd starreduction
sudo apt-get install git-lfs
git lfs pull

Run inference on image.

python removestars.py image_with_stars.jpg  image_without_stars.jpg

Supprots greyscale and RGB images. Alpha channel (if any) in the source image is removed during processing. Gives issues on some types of TIFF files.

Trying out the model in a web browser

Here is link to a online demo of star reduction created using a trained model, TFJS and ReactJS. Please use a desktop browser to access the demo (for memory and performance reasons). The demo runs locally inside your browser, no data outside of your computer. Here is the link to the demo : https://ashishware.com/static/star_removal/index.html