File size: 4,192 Bytes
12d2e9e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 |
"""
Copyright 2021, Dana-Farber Cancer Institute and Weill Cornell Medicine
License: GNU GPL 2.0
"""
import os
import shutil
import urllib
from pathlib import Path
from unittest.mock import MagicMock, patch
import h5py
import numpy as np
import pytest
from pathml.core.utils import writedataframeh5
from pathml.datasets.deepfocus import DeepFocusDataModule, DeepFocusDataset
@pytest.fixture
def create_incomplete_deepfocus_data(tmp_path):
target_dir = tmp_path / "dftests"
target_dir.mkdir(parents=True, exist_ok=True)
filename = target_dir / "outoffocus2017_patches5Classification.h5"
with h5py.File(filename, "w") as f:
X = np.random.randint(low=1, high=254, size=(1000, 64, 64, 3), dtype=np.uint8)
writedataframeh5(f, "X", X)
Y = np.random.randint(low=1, high=5, size=(204000,), dtype=np.uint8)
writedataframeh5(f, "Y", Y)
# Return the path to the file rather than the file object itself to avoid access issues
return filename
def test_incomplete_fails(create_incomplete_deepfocus_data):
target_dir = "dftests"
with pytest.raises(AssertionError):
DeepFocusDataModule(target_dir, download=False)
if os.path.exists(target_dir):
shutil.rmtree(target_dir)
def test_check_deepfocus_data_urls():
# make sure that the urls for the pannuke data are still valid!
url = "https://zenodo.org/record/1134848/files/outoffocus2017_patches5Classification.h5"
r = urllib.request.urlopen(url)
# HTTP status code 200 means "OK"
assert r.getcode() == 200
def check_wrong_path_download_false_fails():
with pytest.raises(AssertionError):
DeepFocusDataModule(data_dir="wrong/path/to/pannuke", download=False)
def test_deepfocusdatamodule_with_incorrect_integrity(create_incomplete_deepfocus_data):
# The `create_incomplete_deepfocus_data` fixture is used to setup the data directory
data_dir = create_incomplete_deepfocus_data
# Attempt to initialize the DeepFocusDataModule with the setup data directory.
# Since the data does not pass the integrity check, it should raise an AssertionError.
with pytest.raises(AssertionError) as excinfo:
DeepFocusDataModule(str(data_dir), batch_size=4, shuffle=False, download=False)
assert (
"download is False but data directory does not exist or md5 checksum failed"
in str(excinfo.value)
)
# TODO: How to test datamodule arguments if checksum without downloading the full dataset?
def create_mock_h5py_file():
"""
Create a mock h5py file with a smaller dataset.
"""
mock_h5py_file = MagicMock()
mock_X = np.random.rand(100, 224, 224, 3) # Smaller image dimensions
mock_Y = np.random.randint(0, 2, size=(100,)) # Binary labels
# Mock the dataset and slicing
mock_h5py_file.__getitem__.side_effect = lambda k: {"X": mock_X, "Y": mock_Y}[k]
return mock_h5py_file
@pytest.mark.parametrize("fold_ix", [1, 2, 3, None])
def test_deepfocus_dataset(fold_ix):
with patch("h5py.File", return_value=create_mock_h5py_file()):
data_dir = Path("fake/path") # Using pathlib.Path for fake data directory
deepfocus_dataset = DeepFocusDataset(data_dir=data_dir, fold_ix=fold_ix)
# Testing data retrieval
img, label = deepfocus_dataset[0]
assert img.shape == (224, 224, 3), "Image shape is incorrect"
assert isinstance(label, np.integer), "Label type is incorrect"
# Additional checks for specific folds
if fold_ix == 1:
# Check if data is from the training set
assert len(deepfocus_dataset) == 100, "Training set size is incorrect"
elif fold_ix == 2:
# Check if data is from the validation set
assert len(deepfocus_dataset) == 0, "Validation set size is incorrect"
elif fold_ix == 3:
# Check if data is from the test set
assert len(deepfocus_dataset) == 0, "Test set size is incorrect"
else:
# If fold_ix is None, it should return the entire dataset
assert (
len(deepfocus_dataset) == 100
), "Dataset size is incorrect for the entire dataset"
|