def reshape_and_perm(x):
x_reshaped = np.reshape(x, [x.shape[0], -1, 3])
x_reshaped = np.reshape(x, [x.shape[0], -1, 3])
channel1 = x_reshaped[:, :, 0]
channel1 = x_reshaped[:, :, 0]
channel1 = np.squeeze(channel1)
print("channel1 shape: {}".format(channel1.shape))
channel1_perm = perm_data(channel1, perm)
channel2 = x_reshaped[:, :, 1]
channel2 = x_reshaped[:, :, 1]
channel2 = np.squeeze(channel2)
print("channel2 shape: {}".format(channel2.shape))
channel2_perm = perm_data(channel2, perm)
channel3 = x_reshaped[:, :, 1]
channel3 = x_reshaped[:, :, 1]
channel3 = np.squeeze(channel3)
print("channel3 shape: {}".format(channel3.shape))
channel3_perm = perm_data(channel3, perm)
stacked_data = np.dstack((channel1_perm, channel2_perm, channel3_perm))
print("stacked data shape: {}".format(stacked_data.shape))
# stacked_data = np.reshape(stacked_data, [stacked_data.shape[0], -1])
return stacked_data
def reshape_and_perm(x):
#testing if removing the sqeeze function fies the error for checking on individual images
x_reshaped = np.reshape(x, [x.shape[0], -1, 3])
#x_reshaped = np.reshape(x, [x.shape[0], -1, 3])
channel1 = x_reshaped[:, :, 0]
#channel1 = x_reshaped[:,:,0]
#channel1 = np.squeeze(channel1)
print("channel1 shape: {}".format(channel1.shape))
channel1_perm = perm_data(channel1, perm)
channel2 = x_reshaped[:, :, 1]
#channel2 = x_reshaped[:,:,1]
#channel2 = np.squeeze(channel2)
print("channel2 shape: {}".format(channel2.shape))
channel2_perm = perm_data(channel2, perm)
channel3 = x_reshaped[:, :, 2]
#channel3 = x_reshaped[:,:,1]
#channel3 = np.squeeze(channel3)
print("channel3 shape: {}".format(channel3.shape))
channel3_perm = perm_data(channel3, perm)
stacked_data = np.dstack((channel1_perm, channel2_perm, channel3_perm))
print("stacked data shape: {}".format(stacked_data.shape))
# stacked_data = np.reshape(stacked_data, [stacked_data.shape[0], -1])
return stacked_data
# Compute coarsened graphs
coarsening_levels = 4
num_vertices, L, perm = coarsen(A, coarsening_levels)
# Compute max eigenvalue of graph Laplacians
lmax = []
for i in range(coarsening_levels):
lmax.append(lmaxX(L[i]))
print('lmax: ' + str([lmax[i] for i in range(coarsening_levels)]))
train_data = tf.Session().run(train_data)
#comment this line for non rotated test images
test_data = tf.Session().run(test_data)
# Reindex nodes to satisfy a binary tree structure
train_data = perm_data(train_data, perm)
val_data = perm_data(val_data, perm)
test_data = perm_data(test_data, perm)
print(train_data.shape)
print(val_data.shape)
print(test_data.shape)
print('Execution time: {:.2f}s'.format(time.time() - t_start))
del perm
class Graph_ConvNet_LeNet5(object):
# Constructor
def __init__(self, net_parameters):