deepwfc3
is a repository of machine learning models built using the Hubble Space Telescope's (HST) Wide Field Camera 3 (WFC3) data. Under projects
, the user will find folders for each model, which includes pretrained weights and biases, data processing scripts, Jupyter Notebooks, and a README.md
briefly explaining the model's purpose. The models have an emphasis on anomaly detection in WFC3 images, which includes IR blobs, UVIS figure 8 ghosts, and more. For more information about WFC3 anomalies, please read WFC3-ISR 2017-22.
Here is a list of the completed models:
- WFC3/IR Blob Classification using Convolutional Neural Networks (CNNs) (WFC3-ISR 2021-08)
- This repository supersedes the original repository.
- WFC3/UVIS Figure 8 Classification using CNNs (WFC3-ISR 2022-03)
In addition, we have some tutorials for implementing more advanced machine learning models in PyTorch and scikit-learn
, our preferred machine learning libraries. Note the tutorials assume the user is familiar with machine learning basic vocabulary and methodology. They DO NOT act as a course for machine learning in general, but as a reference for implementing these models.
Here is a list of the available models with tutorials (all using MNIST data):
- Convolutional Neural Network
- Transfer Learning
- Autoencoders
- Variational Autoencoders
- Dimensionality Reduction with PCA, t-SNE, and UMAP
All the libraries required for using the models are in environment.yml
. The name of the anaconda virtual environment is deepwfc3_env
, which contains standard scientific computing libraries (numpy
, matplotlib
, etc), machine learning frameworks (pytorch
and tensorflow
), and STScI libraries (astropy
, wfc3tools
, etc).
After cloning and changing directories to this repository, create the virtual environment by running this line in a terminal window:
conda env create -f environment.yml
To activate deepwfc3_env
, run this line in a terminal window:
conda activate deepwfc3_env
At the time this was written, the environment uses Python 3.6. It's within our best interest to use the latest software available so we will look into updating our environment sometime in the future.
deepwfc3
follows the Astropy Code of Conduct and strives to provide a
welcoming community to all of our users and contributors.
deepwfc3
is licensed under a 3-clause BSD style license (see the LICENSE.txt
file).