A PyTorch-based tool to generate clouds for satellite images.
The package can now be imported via pip
pip install git+https://github.com/strath-ai/SatelliteCloudGenerator
CloudGenerator()
which encapsulates a specific configuration, and generation probabilities (cloud_p
and shadow_p
). It is compatible as PyTorch module (you can plug it into augmentation pipelines, like torchvision
or albumentations
)my_gen=CloudGenerator(WIDE_CONFIG,cloud_p=1.0,shadow_p=0.5)
my_gen(my_image) # will act just like add_cloud_and_shadow() but will preserve the same configuration!
segmentation_mask(cloud_mask,shadow_mask)
method, which will return a segmentation mask for your generated clouds and shadows!...you can even set a range thin_range
to something like (0.05,0.5)
to also differentiate between thin and thick clouds
and this is an example content of each label:
I hope these features prove useful! :rocket:
min_lvl
and max_lvl
redefined (to refer to cloud strength)add_cloud_and_shadow
locality_index
argument that can control the sparsity of the cloud shape02-Dataset-Generation.ipynb
notebookcloud_hue
Updated DOI, Colab links and citation details.
Pre-release for setting up Zenodo DOI