def __call__(self, inputs):
kspace, mask, target = inputs
# pad if necessary
p1 = max(0, self.shape[0] - target.shape[-2])
p2 = max(0, self.shape[1] - target.shape[-1])
target_padded = torch.nn.functional.pad(
target, (p2 // 2, -(-p2 // 2), p1 // 2, -(-p1 // 2)),
)
# crop if necessary
target_cropped = transforms.center_crop(target_padded, self.shape)
# resimulate
kspace_cropped = transforms.fft2(prep_fft_channel(target_cropped))
new_mask = mask(kspace_cropped.shape).expand_as(kspace_cropped)
new_kspace = unprep_fft_channel(kspace_cropped * new_mask)
new_mask = unprep_fft_channel(new_mask)
tcs = target_cropped.shape[-3]
if not tcs == 2:
target_cropped = target_cropped[
..., ((tcs // 2) // 2) * 2 : ((tcs // 2) // 2) * 2 + 2, :, :
]
return new_kspace, new_mask, target_cropped
def _process_data(self, kspace_data, gt_data):
mask = self.mask_func(kspace_data.shape).expand_as(kspace_data)
kspace_data = unprep_fft_channel(kspace_data.unsqueeze(0))
mask = unprep_fft_channel(mask.unsqueeze(0))
kspace_masked = kspace_data * mask
gt_data = to_complex(gt_data)
if self.keep_mask_as_func:
mask = self.mask_func
return kspace_masked, mask, gt_data
def _process_data(self, kspace_data, gt_data):
mask = self.mask_func(kspace_data.shape, seed=self.seed).expand_as(
kspace_data
)
kspace_data = unprep_fft_channel(kspace_data.unsqueeze(0))
mask = unprep_fft_channel(mask.unsqueeze(0))
kspace_masked = kspace_data * mask
if self.simulate_gt:
gt_data = unprep_fft_channel(gt_data.unsqueeze(0))
if self.keep_mask_as_func:
def mask(shape):
return self.mask_func(shape, seed=self.seed)
return kspace_masked, mask, gt_data
def __call__(self, inputs):
if self.use_target:
tar = inputs[-1]
else:
tar = unprep_fft_channel(
transforms.ifft2(prep_fft_channel(inputs[0])))
norm = torch.norm(tar, p=self.p)
if self.reduction == "mean" and not self.p == "inf":
norm /= np.prod(tar.shape)**(1 / self.p)
if len(inputs) == 2:
return inputs[0] / norm, inputs[1] / norm
else:
return inputs[0] / norm, inputs[1], inputs[2] / norm
)
from reconstruction_methods import admm_l1_rec_diag, grid_search
# ----- load configuration -----
import config # isort:skip
# ------ setup ----------
device = torch.device("cuda:0")
file_name = "grid_search_l1_fourier_"
save_path = os.path.join(config.RESULTS_PATH, "grid_search_l1")
# ----- operators --------
n = (320, 320)
mask_func = RadialMaskFunc(n, 50)
mask = unprep_fft_channel(mask_func((1, 1) + n + (1, )))
OpA = Fourier(mask)
OpTV = TVAnalysisPeriodic(n, device=device)
# ----- load test data --------
test_data_params = {
"mask_func":
mask_func,
"seed":
1,
"filter": [filter_acquisition_no_fs],
"num_sym_slices":
0,
"multi_slice_gt":
False,
"keep_mask_as_func":
RadialMaskFunc,
rotate_real,
unprep_fft_channel,
)
# ----- load configuration -----
import config # isort:skip
# ----- global configuration -----
mpl.use("agg")
device = torch.device("cuda:0")
torch.cuda.set_device(0)
# ----- measurement configuration -----
mask_func = RadialMaskFunc(config.n, 40)
mask = unprep_fft_channel(mask_func((1, 1) + config.n + (1, )))
OpA = Fourier(mask)
inverter = LearnableInverter(config.n, mask, learnable=False)
# ----- network configuration -----
subnet_params = {
"in_channels": 2,
"out_channels": 2,
"drop_factor": 0.0,
"down_blocks": (5, 7, 9, 12, 15),
"up_blocks": (15, 12, 9, 7, 5),
"pool_factors": (2, 2, 2, 2, 2),
"bottleneck_layers": 20,
"growth_rate": 16,
"out_chans_first_conv": 16,
}
def __call__(self, inputs):
kspace, mask, target = inputs
inv = unprep_fft_channel(transforms.ifft2(prep_fft_channel(kspace)))
return inv, target
from reconstruction_methods import admm_l1_rec_diag
# ------ setup ----------
device = torch.device("cuda:0")
torch.cuda.set_device(0)
mask_seed = 123
# ----- operators -----
n = (368, 368)
mask_func = subsample.RandomMaskFunc(
center_fractions=[0.08], accelerations=[4]
)
mask = unprep_fft_channel(
mask_func((1, 1) + n + (1,), seed=mask_seed).expand_as(
torch.ones((1, 1) + n + (1,))
)
)
OpA = Fourier(mask)
OpTV = TVAnalysisPeriodic(n, device=device)
# ----- methods --------
methods = pd.DataFrame(columns=["name", "info", "reconstr", "attacker", "net"])
methods = methods.set_index("name")
noise_ref = noise_gaussian
# ----- set up L1 --------
# grid search parameters for L1 via admm
grid_search_file = os.path.join(
config.RESULTS_PATH, "grid_search_l1", "grid_search_l1_fourier_all.pkl"