Interfaces

Interfaces#

class sdeconv.psfs.interface.SPSFGenerator#

Interface for a psf generator

class sdeconv.deconv.interface.SDeconvFilter#

Interface for a deconvolution filter

All the algorithm settings must be set in the __init__ method (PSF included) and the __call__ method is used to actually do the calculation

class sdeconv.deconv.interface_nn.NNModule#

Deconvolution using the noise to void algorithm

device() → str#

Get the GPU if exists

Returns:

The device name (cuda or CPU)

abstract fit(train_directory: Path, val_directory: Path, n_channel_in: int = 1, n_channels_layer: list[int] = (32, 64, 128), patch_size: int = 32, n_epoch: int = 25, learning_rate: float = 0.001, out_dir: Path = None)#

Train a model on a dataset

Parameters:

  • train_directory – Directory containing the images used for training. One file per image,

  • val_directory – Directory containing the images used for validation of the training. One file per image,

  • n_channel_in – Number of channels in the input images

  • n_channels_layer – Number of channels for each hidden layers of the model,

  • patch_size – Size of square patches used for training the model,

  • n_epoch – Number of epochs,

  • learning_rate – Adam optimizer learning rate

load(filename: Path)#

Load pre-trained model from file

Param:

Path of the model file

save(filename: Path)#

Save the model into file

Param:

Path of the model file

Contents