Dependencies

Third-party software packages required by WaveDiff are installed automatically (see Installation).

Python Dependencies

Package Name	References
numpy	Harris et al. [2020]
scipy	Virtanen et al. [2020]
keras	Chollet and others [2015]
tensorflow	Abadi et al. [2015]
tensorflow-estimator	Abadi et al. [2015]
zernike	Antonello and Verhaegen [2015]
opencv-python	Bradski [2000]
pillow	Clark [2015]
galsim	Rowe et al. [2015]
astropy	Astropy Collaboration et al. [2013], Astropy Collaboration et al. [2018], Astropy Collaboration et al. [2022]
matplotlib	Hunter [2007]
pandas	McKinney and others [2010]
seaborn	Waskom [2021]

Some features in WaveDiff rely on optional third-party packages that are not required for standard training and evaluation workflows.

Package Name	Purpose
tensorflow-addons	Optional optimizers (e.g. RectifiedAdam)

Starting with WaveDiff v3.1.0, tensorflow-addons is no longer a required dependency, as TensorFlow Addons reached end-of-life in May 2024.

By default, WaveDiff uses standard Keras/TensorFlow optimizers (e.g. Adam)
TensorFlow Addons is only imported at runtime if explicitly requested in the configuration
If a TensorFlow Addons optimizer is selected and the package is not installed, WaveDiff will raise a clear runtime error

To use TensorFlow Addons optimizers, install manually:

pip install tensorflow-addons