def __getitem__(self, index):
"""
Return item at index in 0 to len(self)
In this case a set of crops from an lf sample
Return type is a dictionary of depth and colour arrays
"""
with h5py.File(self.file_path, mode='r', libver='latest',
swmr=True) as h5_file:
idx = index // self.num_crops
colour = torch.tensor(h5_file[self.colour][idx],
dtype=torch.float32)
warped = torch.tensor(h5_file[self.warped][idx],
dtype=torch.float32)
grid_size = self.grid_size
sample = {
'colour': colour,
'warped': warped,
'grid_size': grid_size
}
if self.crop_train:
sample = data_transform.get_random_crop(
sample, self.patch_size)
sample = data_transform.normalise_sample(sample)
sample = data_transform.random_gamma(sample)
if self.sub_chan:
sample = data_transform.subsample_channels(sample, 3)
if self.transform:
sample = self.transform(sample)
return sample
def __getitem__(self, index):
"""
Return item at index in 0 to len(self)
In this case a set of crops from an lf sample
Return type is a dictionary of depth and colour arrays
"""
with h5py.File(self.file_path, mode='r', libver='latest',
swmr=True) as h5_file:
idx = index // self.num_crops
depth = torch.squeeze(
torch.tensor(h5_file[self.depth][idx], dtype=torch.float32))
colour = torch.tensor(h5_file[self.colour][idx],
dtype=torch.float32)
grid_size = self.grid_size
sample = {'depth': depth, 'colour': colour, 'grid_size': grid_size}
sample = data_transform.get_random_crop(sample, self.patch_size)
sample = data_transform.random_gamma(sample)
if self.transform:
sample = self.transform(sample)
sample = data_transform.stack(sample, channels=65)
return sample