Photometry of blended galaxies with Deep Learning

Code repository aimed at reprooducing the results presented in the paper.

Set up
How to cite
License

Set up

`candels-blender`

The blend-images used in this analysis have been produced with candels-blender.

Install the code and download the individual galaxies from CANDELS (see instructions)
Choose a seed <SEED> and a total number of blend images <N_BLENDS> and compute a training set and test set
```
candels-blender produce -n <N_BLENDS> --mag_high 23.5 --test_ratio 0.3 --seed <SEED>
```
Prepare the segmentation labels with 3 channels : [overlap, central galaxy, companion galaxy]
```
candels-blender concatenate -d output-s_<SEED>-n_<N_BLENDS> --method ogg_masks
```

Provide a zeropoint to make the flux conversion for the catalog

candels-blender convert -d output-s_<SEED>-n_<N_BLENDS> --zeropoint=25.5

output-s_<SEED>-n_<N_BLENDS> therefore becomes the data directory a.k.a. datadirs

Set the environment variable 'COINBLEND_DATADIR' to your chosen datadir via

export COINBLEND_DATADIR=<path-to-datadir>

`coindeblend`

Clone this repository

git clone https://github.com/aboucaud/coindeblend
cd coindeblend

Install the required dependencies

with conda:

conda env create -f environment.yml
conda activate coindeblend

with pip:

python3 -m pip install -r requirements/requirements.txt

Install coindeblend

python3 -m pip install .

Citing

If you use any of this work, please cite the original publication:

@article{10.1093/mnras/stz3056,
    author = {Boucaud, Alexandre and Huertas-Company, Marc and Heneka, Caroline and Ishida, Emille E O and Sedaghat, Nima and de Souza, Rafael S and Moews, Ben and Dole, Hervé and Castellano, Marco and Merlin, Emiliano and Roscani, Valerio and Tramacere, Andrea and Killedar, Madhura and Trindade, Arlindo M M},
    title = "{Photometry of high-redshift blended galaxies using deep learning}",
    journal = {Monthly Notices of the Royal Astronomical Society},
    year = {2019},
    month = {12},
    issn = {0035-8711},
    doi = {10.1093/mnras/stz3056},
    url = {https://doi.org/10.1093/mnras/stz3056},
    note = {stz3056},
    eprint = {http://oup.prod.sis.lan/mnras/advance-article-pdf/doi/10.1093/mnras/stz3056/31176513/stz3056.pdf},
}

License

The code is published under the BSD 3-Clause License.