def benchmark():
check = True
shape = (40, 41, 42, 150)
affine = np.eye(4)
data = np.ndarray(shape, order="F", dtype=np.float32)
with profile.timestamp("Data_generation"):
data[...] = np.random.standard_normal(data.shape)
target_shape = tuple([s * 1.26 for s in shape[:3]])
target_affine = affine
img = nibabel.Nifti1Image(data, affine)
# Resample one 4D image
if check:
print("Resampling (original)...")
data_orig = utils.timeit(profile(nilearn.resampling_orig.resample_img)
)(img, target_shape=target_shape,
target_affine=target_affine,
interpolation="continuous")
print("Resampling (new)...")
data = utils.timeit(profile(nilearn.resampling.resample_img)
)(img, target_shape=target_shape,
target_affine=target_affine,
interpolation="continuous")
time.sleep(0.5)
del img
time.sleep(0.5)
if check:
np.testing.assert_almost_equal(data_orig.get_data(), data.get_data())
del data
time.sleep(0.5)
def benchmark():
# Concatenate all individual images, time the operation
_, _, images = get_filenames()
if utils.cache_tools_available:
print("Invalidating cache...")
utils.dontneed(images)
print("Concatenating images...")
data = utils.timeit(profile(nilearn.utils.concat_niimgs))(images)
assert(data.shape[3] == len(images))
del data
print("Concatenating images...")
data = utils.timeit(profile(nilearn.utils.concat_niimgs))(images)
assert(data.shape[3] == len(images))
del data