Implementation of Deep Learning unit tests, as well as paths to the 2d data for windows users in the UNet jupyter notebook.

Unit tests have been implemented for classes and functions that relate to Deep Learning tasks such as defining Models, Hyperparameters, Datasets / Dataloaders, as well as the training process.

The process is documented in the following notion file:
https://www.notion.so/qim-dtu/Unit-tests-to-DL-funtions-e8ad1fc1aa134da2ab26474cc6480ca2

the following scripts have been made under qim3d/tests/:

• models/test_unet.py (unit tests for UNet() and Hyperparameters())
• utils/test_augmentations.py (unit tests for Augmentations() and ValueErrors)
• utils/test_data.py (unit tests for Dataset() class, resizing of images with crop, padding, prepare_datasets() and prepare_dataloaders())
• utils/test_models.py (unit tests for output of model_summary() and inference())

Problem:

• Some deep learning tests require considerably more time to run (around 15s to 20s) especially when a Dataloader is involved. This is the case for prepare_dataloaders() as well as the model_summary(). Should they still be included?
• I couldn't think of a way to test the train_model() function, since that requires training a DL model, and would also take too much time.

qim3d/tests/utils/test_data.py

+import qim3d
+import pytest
+
+from torch.utils.data.dataloader import DataLoader
+from qim3d.utils.internal_tools import temp_data
+
+# unit tests for Dataset()
+def test_dataset():
+    img_shape = (32,32)
+    folder = 'folder_data'
+    temp_data(folder, img_shape = img_shape)
+    
+    images = qim3d.utils.Dataset(folder)
+
+    assert images[0][0].shape == img_shape
+
+    temp_data(folder,remove=True)
+
+
+# unit tests for check_resize()
+def test_check_resize():
+    h_adjust,w_adjust = qim3d.utils.data.check_resize(240,240,resize = 'crop',n_channels = 6)
+
+    assert (h_adjust,w_adjust) == (192,192)
+
+def test_check_resize_pad():
+    h_adjust,w_adjust = qim3d.utils.data.check_resize(16,16,resize = 'padding',n_channels = 6)
+
+    assert (h_adjust,w_adjust) == (64,64)
+
+def test_check_resize_fail():
+
+    with pytest.raises(ValueError,match="The size of the image is too small compared to the depth of the UNet. Choose a different 'resize' and/or a smaller model."):
+        h_adjust,w_adjust = qim3d.utils.data.check_resize(16,16,resize = 'crop',n_channels = 6)
+
+
+# unit tests for prepare_datasets()
+def test_prepare_datasets():
+    n = 3
+    validation = 1/3
+    
+    folder = 'folder_data'
+    img = temp_data(folder,n = n)
+
+    my_model = qim3d.models.UNet()
+    my_augmentation = qim3d.utils.Augmentation(transform_test='light')
+    train_set, val_set, test_set = qim3d.utils.prepare_datasets(folder,validation,my_model,my_augmentation)
+
+    assert (len(train_set),len(val_set),len(test_set)) == (int((1-validation)*n), int(n*validation), n)
+
+    temp_data(folder,remove=True)
+
+
+# unit test for validation in prepare_datasets()
+def test_validation():
+    validation = 10
+    
+    with pytest.raises(ValueError,match = "The validation fraction must be a float between 0 and 1."):
+        augment_class = qim3d.utils.prepare_datasets('folder',validation,'my_model','my_augmentation')
+
+
+# unit test for prepare_dataloaders()
+def test_prepare_dataloaders():
+    folder = 'folder_data'
+    temp_data(folder)
+
+    batch_size = 1
+    my_model = qim3d.models.UNet()
+    my_augmentation = qim3d.utils.Augmentation()
+    train_set, val_set, test_set = qim3d.utils.prepare_datasets(folder,1/3,my_model,my_augmentation)
+
+    _,val_loader,_ = qim3d.utils.prepare_dataloaders(train_set,val_set,test_set,
+                                                                           batch_size,num_workers = 1,
+                                                                           pin_memory = False)
+    
+    assert type(val_loader) == DataLoader
+
+    temp_data(folder,remove=True)
\ No newline at end of file