def prep_mnist():
(x_train, y_train), (x_test, y_test) = kd.mnist.load_data()
x = np.array([img.flatten()
for img in np.concatenate((x_train, x_test))]) / 255
y = np.concatenate((y_train, y_test))
xy = np.array([np.append(row, label) for (row, label) in list(zip(x, y))])
print(x.shape, y.shape, xy.shape)
# print(xy[0])
arff_utils.image_stream_to_arff(xy, (28, 28), 'MNIST', 'MNIST.arff')
def prep_mnist_f():
(x_train, y_train), (x_test, y_test) = kd.fashion_mnist.load_data()
x = np.array([img.flatten()
for img in np.concatenate((x_train, x_test))]) / 255
y = np.concatenate((y_train, y_test))
labels = [
'T-shirt', 'Trouser', 'Pullover', 'Dress', 'Coat', 'Sandal', 'Shirt',
'Sneaker', 'Bag', 'Ankle_boots'
]
xy = np.array(
[np.append(row, labels[label]) for (row, label) in list(zip(x, y))])
print(x.shape, y.shape, xy.shape)
# print(xy[0])
arff_utils.image_stream_to_arff(xy, (28, 28), 'MNIST_F', 'MNIST_F.arff')
def prep_cmater(root):
path = f'{root}/raw-batch/vis/cmater/datasets/bangla-numerals/training-images.npz'
data = np.load(path)
x = np.array([[[np.average(c) / 255 for c in row] for row in img]
for img in data.f.images])
x = np.array([img.flatten() for img in x])
y = data.f.labels
xy = np.array([
np.append(row, label) for _ in range(2)
for (row, label) in list(zip(x, y))
])
np.random.shuffle(xy)
print(x.shape, y.shape, xy.shape)
arff_utils.image_stream_to_arff(xy, (32, 32), 'CMATER-BANGLA',
'CMATER-BANGLA.arff')
def prep_cifar():
(x_train, y_train), (x_test, y_test) = kd.cifar10.load_data()
x_rgb = np.concatenate((x_train, x_test))
x = np.array([[[np.average(c) / 255 for c in row] for row in img]
for img in x_rgb])
x = np.array([img.flatten() for img in x])
y = np.concatenate((y_train, y_test)).flatten()
labels = [
'airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog',
'horse', 'ship', 'truck'
]
xy = np.array(
[np.append(row, labels[label]) for (row, label) in list(zip(x, y))])
print(x.shape, y.shape, xy.shape)
# print(x_train[1][0][1], xy[1][1])
arff_utils.image_stream_to_arff(xy, (32, 32), 'CIFAR10', 'CIFAR10.arff')
def prep_imagenette(root):
path = f'{root}/raw-batch/vis/imagenette/train_64'
xy = []
for label in os.listdir(path):
for image_file_path in glob.glob(path + '/' + label + '/*'):
try:
img = imread(image_file_path)
x = np.array([[np.average(c) / 255 for c in row]
for row in img]).flatten()
y = label
xy.append(np.append(x, y))
print(image_file_path, x.shape, y)
except ValueError as e:
print(f'Could not read {image_file_path}: {e}')
xy = np.array(xy)
np.random.shuffle(xy)
print(xy.shape)
arff_utils.image_stream_to_arff(xy, (64, 64), 'IMAGENETTE',
'IMAGENETTE.arff')
def prep_malaria(root):
pos_path = f'{root}/raw-batch/vis/malaria/32/Parasitized'
neg_path = f'{root}/raw-batch/vis/malaria/32/Uninfected'
train_files = glob.glob(pos_path + "/*") + glob.glob(neg_path + "/*")
xy = []
for i, name in enumerate(train_files):
try:
img = imread(name)
y = 'pos' if 'Parasitized' in name else 'neg'
x = np.array([[np.average(c) / 255 for c in row]
for row in img]).flatten()
xy.append(np.append(x, y))
print(i, name, x.shape, y)
except ValueError as e:
print(f'Could not read {name}: {e}')
xy = np.array(xy)
np.random.shuffle(xy)
print(xy.shape)
arff_utils.image_stream_to_arff(xy, (32, 32), 'MALARIA', 'MALARIA.arff')
def prep_intel(root):
path = f'{root}/raw-batch/vis/intel_imgs/seg_train_32'
xy = []
for label in os.listdir(path):
for image_file_path in glob.glob(path + '/' + label + '/*'):
try:
img = imread(image_file_path)
x = np.array([[np.average(c) / 255 for c in row]
for row in img]).flatten()
y = label
for _ in range(2):
xy.append(np.append(x, y))
print(image_file_path, x.shape, y)
except ValueError as e:
print(f'Could not read {image_file_path}: {e}')
xy = np.array(xy)
np.random.shuffle(xy)
print(xy.shape)
arff_utils.image_stream_to_arff(xy, (32, 32), 'INTEL-IMGS',
'INTEL-IMGS.arff')
def prep_dogs_vs_cats(root):
path = f'{root}/raw-batch/vis/dogs-vs-cats/train_32'
train_files = glob.glob(path + "/*")
xy = []
for i, name in enumerate(train_files):
try:
img = imread(name)
y = 'cat' if 'cat' in name.split('/')[-1] else 'dog'
x = np.array([[np.average(c) / 255 for c in row]
for row in img]).flatten()
xy.append(np.append(x, y))
print(i, name, x.shape, y)
except ValueError as e:
print(f'Could not read {name}: {e}')
xy = np.array(xy)
np.random.shuffle(xy)
print(xy.shape)
arff_utils.image_stream_to_arff(xy, (32, 32), 'DOGS-VS-CATS',
'DOGS-VS-CATS.arff')