This package provides the tools necessary to train a neural network to denoise synchrotron x-ray area detector images.
Source code | https://github.com/garryod/ad_denoise |
Documentation | https://garryod.github.io/ad_denoise |
Releases | https://github.com/garryod/ad_denoise/releases |
Model training can be performed using the command below:
python -m ad_denoise train my_config.yaml
Where my_config.yaml
is as so:
max_epochs: 50
model:
Noise2Self:
network:
Gaussian:
kernel_half_width: 3
train_dataset:
Hdf5ADImagesDataset:
data_paths:
- /dls/i22/data/2022/cm31149-3/Denoising/i22-629817.nxs
frame_key: entry1/detector/data
count_times_key: entry1/instrument/detector/count_time
mask_path: /dls/i22/data/2022/cm31149-3/processing/SAXS_mask.nxs
mask_key: entry/mask/mask
val_dataset:
InputTargetDataset:
input:
Hdf5ADImagesDataset:
data_paths:
- /dls/i22/data/2022/cm31149-3/Denoising/i22-629817.nxs
frame_key: entry1/detector/data
count_times_key: entry1/instrument/detector/count_time
mask_path: /dls/i22/data/2022/cm31149-3/processing/SAXS_mask.nxs
mask_key: entry/mask/mask
target:
Hdf5ADImagesDataset:
data_paths:
- /dls/i22/data/2022/cm31149-3/Denoising/i22-629822.nxs
- /dls/i22/data/2022/cm31149-3/Denoising/i22-629823.nxs
frame_key: entry1/detector/data
count_times_key: entry1/instrument/detector/count_time
mask_path: /dls/i22/data/2022/cm31149-3/processing/SAXS_mask.nxs
mask_key: entry/mask/mask
See https://garryod.github.io/ad_denoise for more detailed documentation.