|
|
|
|
|
|
|
""" |
|
|
|
import os |
|
import numpy as np |
|
from matplotlib import pyplot as plt |
|
import cv2 |
|
import keras |
|
from keras.models import Sequential |
|
from keras.layers import Dense, Dropout |
|
from tensorflow.keras.utils import to_categorical |
|
|
|
""" |
|
|
|
def upload_data(path_name, number_of_class, number_of_images): |
|
X_Data = [] |
|
Y_Data = [] |
|
for i in range(number_of_class): |
|
images = os.listdir(path_name + str(i)) |
|
for j in range(number_of_images): |
|
img = cv2.imread(path_name + str(i)+ '/' + images[j], 0) |
|
X_Data.append(img) |
|
Y_Data.append(i) |
|
print("> the " + str(i) + "-th file is successfully uploaded.", end='\r') |
|
return np.array(X_Data), np.array(Y_Data) |
|
|
|
n_class = 33 |
|
n_train = 2000 |
|
n_test = 500 |
|
|
|
x_train, y_train = upload_data('/media/etabook/etadisk1/EducFils/PFE/DATA2/train_data/', n_class, n_train) |
|
x_test, y_test = upload_data('/media/etabook/etadisk1/EducFils/PFE/DATA2/test_data/', n_class, n_test) |
|
|
|
|
|
print("The x_train's shape is :", x_train.shape) |
|
print("The x_test's shape is :", x_test.shape) |
|
print("The y_train's shape is :", y_train.shape) |
|
print("The y_test's shape is :", y_test.shape) |
|
|
|
def plot_data(num=3): |
|
fig, axes = plt.subplots(1, num, figsize=(12, 8)) |
|
for i in range(num): |
|
index = np.random.randint(len(x_test)) |
|
axes[i].imshow(np.reshape(x_test[index], (28, 28))) |
|
axes[i].set_title('image label: %d' % y_test[index]) |
|
axes[i].axis('off') |
|
|
|
plt.show() |
|
|
|
plot_data(num=5) |
|
|
|
plot_data(num=5) |
|
|
|
num_classes = 33 |
|
size = 28 |
|
|
|
x_train = x_train.astype('float32') |
|
x_test = x_test.astype('float32') |
|
x_train = np.reshape(x_train, (x_train.shape[0], size*size)) |
|
x_test = np.reshape(x_test, (x_test.shape[0], size*size)) |
|
x_train /= 255 |
|
x_test /= 255 |
|
print('x_train shape:', x_train.shape) |
|
print(x_train.shape[0], 'train samples') |
|
print(x_test.shape[0], 'test samples') |
|
|
|
|
|
y_train = to_categorical(y_train, num_classes) |
|
y_test = to_categorical(y_test, num_classes) |
|
|
|
"""## Define our neural network model (Architecture)""" |
|
|
|
model = Sequential() |
|
model.add(Dense(512, input_shape=(size*size,), activation='relu')) |
|
model.add(Dense(128, activation='relu')) |
|
model.add(Dropout(0.3)) |
|
model.add(Dense(num_classes, activation='softmax')) |
|
|
|
model.compile(loss=keras.losses.categorical_crossentropy, |
|
metrics=['accuracy']) |
|
|
|
model.summary() |
|
|
|
"""## Model prediction on test data before training """ |
|
|
|
def plot_predictions(model, num=3): |
|
fig, axes = plt.subplots(1, num, figsize=(12, 8)) |
|
for i in range(num): |
|
index = np.random.randint(len(x_test)) |
|
pred = np.argmax(model.predict(np.reshape(x_test[index], (1, size*size)))) |
|
axes[i].imshow(np.reshape(x_test[index], (size, size))) |
|
axes[i].set_title('Predicted label: '+ str(pred) + '\n/ true label :'+ str([e for e, x in enumerate(y_test[index]) if x == 1][0])) |
|
axes[i].axis('off') |
|
|
|
plt.show() |
|
|
|
plot_predictions(model, num=5) |
|
|
|
"""## Training""" |
|
|
|
history = model.fit(x_train, y_train, batch_size=128, epochs=20, validation_data=(x_test, y_test)) |
|
|
|
"""## Model prediction on test data after training""" |
|
|
|
plot_predictions(model, num=5) |
|
score = model.evaluate(x_test, y_test, verbose = 0) |
|
print('Test loss:', score[0]) |
|
print('Test accuracy:', score[1]) |
|
|
|
"""## Model history during training""" |
|
|
|
import matplotlib.pyplot as plt |
|
import numpy as np |
|
with plt.xkcd(): |
|
plt.plot(history.history['accuracy'], color='c') |
|
plt.plot(history.history['val_accuracy'], color='red') |
|
plt.title('Tifinagh-MNIST model accuracy') |
|
plt.legend(['acc', 'val_acc']) |
|
plt.savefig('acc_Tifinagh_MNIST.png') |
|
plt.show() |
|
|
|
with plt.xkcd(): |
|
plt.plot(history.history['loss'], color='c') |
|
plt.plot(history.history['val_loss'], color='red') |
|
plt.title('Tifinagh-MNIST model loss') |
|
plt.legend(['loss', 'val_loss']) |
|
plt.savefig('loss_Tifinagh_MNIST.png') |
|
plt.show() |