This repo generates the training data and the model for Morpheus-Deblend.

This is the active development repo for the project and as such you will find many versions and experiments for different approaches to the problem.

I recommend you work in a virtual environment.

• astropy
• comet_ml
• flow_vis
• gin
• numpy
• requests
• tensorflow==2.2
• tensorflow_addons
• scarlet (https://github.com/pmelchior/scarlet)
• scipy
• sharedmem
• scikit-image
• scikit-learn
• tqdm

After the requirements are installed install the local repo:

pip install -e .

You can then download the data by running

python src/data/make_dataset.py

Then generate the training data using

python src/features/build_features.py

Then train a model by running

python src/models/train_model.py src/models/morpheus-deblend.gin

In order to run the training you need to have a comet.ml account setup as that is where the training is logged.