pca_all = utils.PCA(D=train_data_all_flat,
n_components=train_data_all_flat.shape[1])
logger.info('Whitening data...')
train_data_all_flat = pca_all.transform(D=train_data_all_flat,
whiten=True,
ZCA=True)
train_data_all = train_data_all_flat.T.reshape(train_data_all.shape[0:2] +
(-1, ))
test_data_flat = pca_all.transform(D=test_data_flat, whiten=True, ZCA=True)
test_data = test_data_flat.T.reshape(test_data.shape[0:2] + (-1, ))
# Datasets
train_set = utils.NumpyDataset(train_data_all,
train_labels_all,
transform=utils.NumpyToTensor())
val_set = utils.NumpyDataset(test_data,
test_labels,
transform=utils.NumpyToTensor())
test_set = utils.NumpyDataset(test_data,
test_labels,
transform=utils.NumpyToTensor())
# Data loaders
train_loader = DataLoader(train_set,
batch_size=batch_size,
shuffle=True,
num_workers=8)
val_loader = DataLoader(val_set,
batch_size=batch_size,
shuffle=False,
# load square aid
f = h5py.File(data_directory + '/aid' + str(resize_size[0]) + '.h5', 'r')
data = f['images'][()]
labels = f['labels'][()]
class_names = f.attrs['class_names']
f.close()
# Normalize data
data = utils.normalize(data)
data = data.reshape(data.shape[0], data.shape[1], resize_size[0],
resize_size[1])
# Datasets
dataset = utils.NumpyDataset(data, labels, transform=utils.NumpyToTensor())
# Run the experiments
for seed in seeds:
# Data loaders
logger.info('Split data with seed {}'.format(seed))
torch.manual_seed(seed)
np.random.seed(seed)
train_indices = []
val_indices = []
for cls in np.unique(labels):
indices = np.where(labels == cls)
indices = np.random.permutation(indices[0])
train_indices.append(
indices[:int(len(indices) * (1 - validating_ratio))])
val_indices.append(
data = f['images'][()]
labels = f['labels'][()]
index_matrix = torch.tensor(f['index_matrix'][()])
class_names = f.attrs['class_names']
f.close()
data_shifted = np.zeros(data.shape[0:2] + index_matrix.shape).astype(
np.float32)
for i in range(index_matrix.shape[0]):
for j in range(index_matrix.shape[1]):
if not int(index_matrix[i, j]) == -1:
data_shifted[:, :, i, j] = data[:, :,
int(index_matrix[i, j])]
sh_dataset = utils.NumpyDataset(data_shifted,
labels,
transform=utils.NumpyToTensor())
hex_dataset = utils.NumpyDataset(data,
labels,
transform=utils.NumpyToTensor())
sh_loader = DataLoader(sh_dataset,
batch_size=batch_size,
shuffle=False)
hex_loader = DataLoader(hex_dataset,
batch_size=batch_size,
shuffle=False)
logger.info('batch size: {} iterations: {}'.format(
batch_size, len(hex_loader)))
index_matrix = index_matrix.unsqueeze_(0).unsqueeze_(0)
