if K.image_data_format() == 'channels_first':
input_size = (3,71,71)
else:
input_size = (71,71,3)
output_size = 5
handler = data_utils.data_handler(data_dir_path, sample_fractions=sample_fractions,
input_size=input_size, labels_type='classes',
output_size=output_size, normalize_input=False,
create_samples_bool=False, preprocess_bool=False,
crp_factor=2, ds_factor=3)
### Load data
X_train, y_train = data_utils.load_samples(handler, 'training', grey_scale=False, as_vector=False)
X_val, y_val = data_utils.load_samples(handler, 'validation', grey_scale=False, as_vector=False)
### Convert class vectors to binary class matrices
y_train = np_utils.to_categorical(y_train, n_classes)
y_val = np_utils.to_categorical(y_val, n_classes)
### Build model
model = conv_net.CNN_VGG16(input_size=input_size, output_size=n_classes)
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
#plot_model(model, to_file='model_plot.png', show_shapes=True, show_layer_names=True)
# this is the augmentation configuration for training
train_datagen = ImageDataGenerator(
rescale=1./255,
rotation_range=90,
classNames = {0: 'Disc',
1: 'Spiral',
2: 'Elliptical',
3: 'Round',
4: 'Other'}
n_comp = 200
handler = data_utils.data_handler(data_dir_path, sample_fractions=sample_fractions,
input_size=input_size, labels_type='classes',
output_size=output_size, normalize_input=False,
create_samples_bool=False, preprocess_bool=False,
crp_factor=2, ds_factor=3)
### Load data
X_train, y_train = data_utils.load_samples(handler, 'training', grey_scale=True)
X_val, y_val = data_utils.load_samples(handler, 'validation', grey_scale=True)
### Perform PCA
X_train_pca, X_val_pca, pca = data_utils.compute_pca(
X_train=X_train, n_comp=n_comp, X_test=X_val)
### Train an one vs. all Logistic Regression model
param_grid = {'C': 10.**np.arange(-8, 8, 0.5)}
clf = GridSearchCV(LogisticRegression(multi_class='ovr'), param_grid, scoring="accuracy")
t0 = time()
clf = clf.fit(X_train_pca, y_train)
print("done in %0.3fs" % (time() - t0))
print(clf.best_estimator_)
input_size = (71, 71)
output_size = 1
handler = data_utils.data_handler(data_dir_path,
sample_fractions=sample_fractions,
input_size=input_size,
labels_type='classes',
output_size=output_size,
normalize_input=False,
create_samples_bool=False,
preprocess_bool=False,
crp_factor=2,
ds_factor=3)
X_train, y_train = data_utils.load_samples(handler,
sample='training',
grey_scale=True)
# PCA Variance Explained
X_train_pca, pca = data_utils.compute_pca(X_train, n_comp=X_train.shape[1])
plt.plot(np.cumsum(pca.explained_variance_ratio_))
plt.xlabel('Number of Components')
plt.ylabel('Explained Variance')
plt.title('Cumulative Variance Explained ')
plt.show()
plt.savefig(os.path.join(figures_dir_path, 'PCA_Variance_Explained.png'))
# Reconstruted Images
raw_img = X_train[5000, :].reshape(input_size)
vector_img = raw_img.reshape(np.prod(raw_img.shape))