def test_vid4_dataset():
root_path = Path(__file__).parent / 'data'
txt_content = ('calendar 1 (320,480,3)\ncity 2 (320,480,3)\n')
mocked_open_function = mock_open(read_data=txt_content)
with patch('builtins.open', mocked_open_function):
vid4_dataset = SRVid4Dataset(lq_folder=root_path / 'lq',
gt_folder=root_path / 'gt',
ann_file='fake_ann_file',
num_input_frames=5,
pipeline=[],
scale=4,
test_mode=False,
metric_average_mode='clip',
filename_tmpl='{:08d}')
assert vid4_dataset.data_infos == [
dict(lq_path=str(root_path / 'lq'),
gt_path=str(root_path / 'gt'),
key='calendar/00000000',
num_input_frames=5,
max_frame_num=1),
dict(lq_path=str(root_path / 'lq'),
gt_path=str(root_path / 'gt'),
key='city/00000000',
num_input_frames=5,
max_frame_num=2),
dict(lq_path=str(root_path / 'lq'),
gt_path=str(root_path / 'gt'),
key='city/00000001',
num_input_frames=5,
max_frame_num=2),
]
with pytest.raises(AssertionError):
# num_input_frames should be odd numbers
SRVid4Dataset(lq_folder=root_path,
gt_folder=root_path,
ann_file='fake_ann_file',
num_input_frames=6,
pipeline=[],
scale=4,
test_mode=False)
with pytest.raises(ValueError):
# metric_average_mode can only be either 'clip' or 'all'
SRVid4Dataset(lq_folder=root_path,
gt_folder=root_path,
ann_file='fake_ann_file',
num_input_frames=6,
pipeline=[],
scale=4,
metric_average_mode='abc',
test_mode=False)
def test_vid4_dataset():
root_path = Path(__file__).parent.parent.parent / 'data'
txt_content = ('calendar 1 (320,480,3)\ncity 2 (320,480,3)\n')
mocked_open_function = mock_open(read_data=txt_content)
with patch('builtins.open', mocked_open_function):
vid4_dataset = SRVid4Dataset(
lq_folder=root_path / 'lq',
gt_folder=root_path / 'gt',
ann_file='fake_ann_file',
num_input_frames=5,
pipeline=[],
scale=4,
test_mode=False,
metric_average_mode='clip',
filename_tmpl='{:08d}')
assert vid4_dataset.data_infos == [
dict(
lq_path=str(root_path / 'lq'),
gt_path=str(root_path / 'gt'),
key='calendar/00000000',
num_input_frames=5,
max_frame_num=1),
dict(
lq_path=str(root_path / 'lq'),
gt_path=str(root_path / 'gt'),
key='city/00000000',
num_input_frames=5,
max_frame_num=2),
dict(
lq_path=str(root_path / 'lq'),
gt_path=str(root_path / 'gt'),
key='city/00000001',
num_input_frames=5,
max_frame_num=2),
]
# test evaluate function ('clip' mode)
results = [{
'eval_result': {
'PSNR': 21,
'SSIM': 0.75
}
}, {
'eval_result': {
'PSNR': 22,
'SSIM': 0.8
}
}, {
'eval_result': {
'PSNR': 24,
'SSIM': 0.9
}
}]
eval_results = vid4_dataset.evaluate(results)
np.testing.assert_almost_equal(eval_results['PSNR'], 22)
np.testing.assert_almost_equal(eval_results['SSIM'], 0.8)
# test evaluate function ('all' mode)
vid4_dataset = SRVid4Dataset(
lq_folder=root_path / 'lq',
gt_folder=root_path / 'gt',
ann_file='fake_ann_file',
num_input_frames=5,
pipeline=[],
scale=4,
test_mode=False,
metric_average_mode='all',
filename_tmpl='{:08d}')
eval_results = vid4_dataset.evaluate(results)
np.testing.assert_almost_equal(eval_results['PSNR'], 22.3333333)
np.testing.assert_almost_equal(eval_results['SSIM'], 0.81666666)
with pytest.raises(AssertionError):
# num_input_frames should be odd numbers
SRVid4Dataset(
lq_folder=root_path,
gt_folder=root_path,
ann_file='fake_ann_file',
num_input_frames=6,
pipeline=[],
scale=4,
test_mode=False)
with pytest.raises(ValueError):
# metric_average_mode can only be either 'folder' or 'all'
SRVid4Dataset(
lq_folder=root_path,
gt_folder=root_path,
ann_file='fake_ann_file',
num_input_frames=5,
pipeline=[],
scale=4,
metric_average_mode='abc',
test_mode=False)
with pytest.raises(TypeError):
# results must be a list
vid4_dataset.evaluate(results=5)
with pytest.raises(AssertionError):
# The length of results should be equal to the dataset len
vid4_dataset.evaluate(results=[results[0]])