deepmreye.util package

Submodules

deepmreye.util.data_generator module

deepmreye.util.data_generator.create_cv_generators(dataset, num_cvs=5, **args)

deepmreye.util.data_generator.create_generators(full_training_list, full_testing_list, batch_size=8, withinsubject_split=None, augment_list=None, mixed_batches=True, inner_timesteps=None)

deepmreye.util.data_generator.create_holdout_generators(datasets, train_split=0.6, **args)

deepmreye.util.data_generator.create_leaveoneout_generators(datasets, training_subset=None, **args)

deepmreye.util.data_generator.data_generator(file_list, batch_size, training=False, mixed_batches=True, withinsubject_split=None, augment_list=None, inner_timesteps=None)

Take a random subject, load it and return a batched subset.

deepmreye.util.data_generator.get_all_subject_data(fn_subject)

deepmreye.util.data_generator.get_nonan_indices(file_list)

deepmreye.util.data_generator.get_single_data_generators(fn_list, batch_size, string_cut=4, **args)

deepmreye.util.data_generator.get_start_end_tr(withinsubject_split, training)

deepmreye.util.data_generator.get_subject_data(fn_subject, batch_size=None, sample_index=None, start_tr=None, end_tr=None, nonan_indices=None)

deepmreye.util.data_generator.get_tr_indices(num_trs, start_tr, end_tr)

deepmreye.util.data_io module

deepmreye.util.data_io.download_mask(data_path, remote_path='https://github.com/DeepMReye/DeepMReye/blob/main/deepmreye/masks/')

deepmreye.util.data_io.get_all_subject_labels(subject_string, mat_data, num_downsampled=10, use_real=False)

For models with multiple outputs we want to estimate the sub-TR XY.

Inputs:

• subject_string : Subject identified

• mat_data : Data to subject logs in mat format

• num_downsampled : How many sub-TR XY are left in the output

deepmreye.util.data_io.get_all_subject_labels_bmd(subject_string, run_idx, num_downsampled=10, real_et=False)

For models with multiple outputs we want to estimate the sub-TR XY.

Parameters

subject_string :

Subject identified

run_idx :

Index for run

num_downsampled :

How many sub-TR XY are left in the output

deepmreye.util.data_io.get_all_subject_labels_ign(subject_string, num_downsampled=10)

For models with multiple outputs we want to estimate the sub-TR XY.

Parameters

subject_string :

Subject identified

mat_data :

Data to subject logs in mat format

num_downsampled :

How many sub-TR XY are left in the output

deepmreye.util.data_io.get_all_subject_labels_mmd(subject_string, run_idx, num_downsampled=10)

For models with multiple outputs we want to estimate the sub-TR XY.

Parameters

subject_string :

Subject identified

run_idx :

Index for run

num_downsampled :

How many sub-TR XY are left in the output

deepmreye.util.data_io.get_subject_labels(subject_string, mat_data)

deepmreye.util.model_opts module

deepmreye.util.model_opts.get_opts()

deepmreye.util.util module

Additional methods which did not earn its own space in the main methods.

Maybe because they are more general and made for higher purposes.

class deepmreye.util.util.Arg(*args, **kwargs)

Bases: object

deepmreye.util.util.angle_between_points(y_true, y_pred)

deepmreye.util.util.augment_input(X, rotation=0, shift=0, zoom=0)

Augment 3D images.

Parameters

X :

Batch of 3D images

rotation :

Rotation in degree

shift :

Shift in pixels

zoom :

Zoom in factor

Returns

X :

Augmented batch of 3D images

deepmreye.util.util.calculate_scores(y_true, y_pred, euc_pred, percentile_cut=None)

class deepmreye.util.util.color

Bases: object

BLUE = '\x1b[94m'

BOLD = '\x1b[1m'

CYAN = '\x1b[96m'

DARKCYAN = '\x1b[36m'

END = '\x1b[0m'

GREEN = '\x1b[92m'

PURPLE = '\x1b[95m'

RED = '\x1b[91m'

UNDERLINE = '\x1b[4m'

YELLOW = '\x1b[93m'

deepmreye.util.util.euclidean_distance(y_true, y_pred)

deepmreye.util.util.get_model_scores(real_y, pred_y, euc_pred, **args)

deepmreye.util.util.mish(x)

deepmreye.util.util.quantify_predictions(y_true, y_pred, euc_pred, subtr_functor=<function median>, percentile_cut=None)

deepmreye.util.util.smooth_signal(signal, N)

Smooth by convolving a filter with 1/N.

Parameters

signalarray_like

Signal to be smoothed

Nint

smoothing_factor

Returns

signalarray_like

Smoothed signal

deepmreye.util.util.step_decay_schedule(initial_lr=0.0001, decay_factor=0.9, num_epochs=50)

Module contents