from perclearn import mnist_reader
from perclearn.utils import (
create_2D_noise,
scale_2D,
create_new_dataset,
)
cwd = os.getcwd()
X_train, y_train = mnist_reader.load_mnist(opj(cwd, 'perclearn/data/fashion'),
kind='train')
X_test, y_test = mnist_reader.load_mnist(opj(cwd, 'perclearn/data/fashion'),
kind='t10k')
X_train = create_new_dataset(X_train, [[0, 0]])
np.savez(opj(cwd, 'perclearn/data/experiments/1/training'), X_train)
X_test = create_new_dataset(X_test, [[0, 28], [28, 0], [28, 28]])
np.savez(opj(cwd, 'perclearn/data/experiments/1/testing'), X_test)
dataset = X_test
i = 10
offsets = [[0, 28], [28, 0], [28, 28]]
m, n = dataset.shape
im_x = int(np.sqrt(n))
num_offsets = len(offsets)
new_dataset = np.zeros((m, n * 4))
model.fit(x_train,
y_train,
validation_data=(x_val, y_val),
batch_size=batch_size,
epochs=epochs,
verbose=1)
"""
TRANSLATION
"""
score = np.empty((2, 28))
# iterate over creation of different x_tests and evaluate model
for offset_x in range(28):
x_test, _ = create_new_dataset(x_test_full, [[offset_x, 0]])
if K.image_data_format() == 'channels_first':
x_test = x_test.reshape(x_test.shape[0], 1, img_rows, img_cols)
else:
x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)
x_test = x_test.astype('float32')
x_test /= 255
print(x_test.shape[0], 'test samples')
print('offset x: ', offset_x)
score[:, offset_x] = np.array(model.evaluate(x_test, y_test, verbose=0))
print('Test loss:', score[0, offset_x])
print('Test accuracy:', score[1, offset_x])