def get_tr_te_set():
"""
This function will extract dlib feature and face shape labels for initial cartoon dataset (10000) for training and
additional cartoon dataset (2500) for testing.
:return:
features: an np array containing flatten 17 landmarks features for training
features_te: an np array containing flatten 17 landmarks features for test
labels: an list containing labels for training set
labels_te: an list containing labels for test set
"""
# extracting dlib features and gender labels from training set and test set
print("Extraction begin")
features, labels = ex_b1.extract_features_labels()
features_te, labels_te = ex_te_b1.extract_features_labels()
print("Extraction end")
# transfer to numpy array to do matrix operation
features = np.array(features)
features_te = np.array(features_te)
# flatten data for training
features = features.reshape(features.shape[0], 17 * 2)
features_te = features_te.reshape(features_te.shape[0], 17 * 2)
# return features and labels for training and testing
return features, features_te, labels, labels_te
def get_tr_te_set():
"""
This function will extract dlib feature and labels for all CelebA images.
It also provided prepared training, validation and test set with ratio 6:2:2
:return:
features_tr: an array containing flatten 68 landmarks features for training
features_vali: an array containing flatten 68 landmarks features for validation
features_te: an array containing flatten 68 landmarks features for test
labels_tr: an list containing labels for training set
labels_vali: an list containing labels for validation set
labels_te: anlist containing labels for test set
"""
print("Extraction begin")
features, labels = ex.extract_features_labels()
features_te, labels_te = ex_te.extract_features_labels()
print("Extraction end")
features = np.array(features)
features_te = np.array(features_te)
print(features.shape)
print(features_te.shape)
features_tr = features[:features_te.shape[0] * 3]
features_vali = features[features_te.shape[0] * 3:features_te.shape[0] * 4]
labels_tr = labels[:features_te.shape[0] * 3]
labels_vali = labels[features_te.shape[0] * 3:features_te.shape[0] * 4]
features_tr = features_tr.reshape(features_te.shape[0] * 3, 17 * 2)
features_vali = features_vali.reshape(features_te.shape[0], 17 * 2)
features_te = features_te.reshape(features_te.shape[0], 17 * 2)
return features_tr, features_vali, features_te, labels_tr, labels_vali, labels_te
def get_tr_te_set():
"""
This function will automatically load the images
inside dataset with given train and test set number
:param num_tr: number of train set
:param num_te: number of test set
:return: train set and test set
"""
print("Extraction begin")
features, labels = ex.extract_features_labels()
features_te, labels_te = ex_te.extract_features_labels()
print("Extraction end")
features = np.array(features)
#labels = np_utils.to_categorical(labels, 5)
features_te = np.array(features_te)
#labels_te = np_utils.to_categorical(labels_te, 5)
features_tr = features[:features_te.shape[0] * 3]
features_vali = features[features_te.shape[0] * 3:features_te.shape[0] * 4]
labels_tr = labels[:features_te.shape[0] * 3]
labels_vali = labels[features_te.shape[0] * 3:features_te.shape[0] * 4]
features_tr = features_tr.reshape(features_te.shape[0] * 3, 17 * 2)
features_vali = features_vali.reshape(features_te.shape[0], 17 * 2)
features_te = features_te.reshape(features_te.shape[0], 17 * 2)
return features_tr, features_vali, features_te, labels_tr, labels_vali, labels_te