netsatsawat’s gists

netsatsawat / model_predict.py

Created June 18, 2019 06:27

Use the network to predict the test data generator

	STEP_SIZE_TEST = testing_generator.n // testing_generator.batch_size
	testing_generator.reset()
	predictions = malaria_model.predict_generator(testing_generator,
	steps=STEP_SIZE_TEST,
	verbose=1)
	img_name = testing_generator.filenames
	results = pd.DataFrame({'img_name': img_name,
	'prediction': predicted_class})
	_this = results.img_name.str.split('\\', n=1, expand=True)
	results['reference'] = _this[0]

netsatsawat / train_network.py

Created June 18, 2019 06:16

Train the malaria model

	sess = K.get_session()
	init = tf.global_variables_initializer()
	sess.run(init)
	STEP_SIZE_TRAIN = training_generator.samples // training_generator.batch_size
	STEP_SIZE_VALID = validatation_generator.n // validatation_generator.batch_size
	training_stats = CollectStats()
	history = malaria_model.fit_generator(training_generator,
	epochs=50,
	steps_per_epoch=STEP_SIZE_TRAIN,
	validation_data=validatation_generator,

netsatsawat / create_top_classifier_layer.py

Created June 18, 2019 06:09

Create top layer

	malaria_model = tf.keras.Sequential([features_extractor_layer,
	layers.Dense(training_generator.num_classes,
	activation='softmax')
	])
	malaria_model.summary()

	malaria_model.compile(optimizer=tf.keras.optimizers.RMSprop(lr=0.001),
	loss='categorical_crossentropy',
	metrics=['acc'])
	"""

netsatsawat / data_gen.py

Created June 18, 2019 05:53

Use to demonstrate the ImageDataGenerator class and flow_from_directory() method

	training_datagen = tf.keras.preprocessing.image.ImageDataGenerator(rescale=1./255.,
	rotation_range=30,
	zoom_range=0.3,
	width_shift_range=0.2,
	height_shift_range=0.2,
	fill_mode='nearest',
	horizontal_flip=True)
	validation_datagen = tf.keras.preprocessing.image.ImageDataGenerator(rescale=1./255.)
	testing_datagen = tf.keras.preprocessing.image.ImageDataGenerator(rescale=1./255.)
	training_generator = training_datagen.flow_from_directory('../data/Train/',

netsatsawat / load_pretrain.py

Created June 18, 2019 04:20

For loading pretrained feature extractor - Malaria dataset

	feature_extractor_url = "https://tfhub.dev/google/imagenet/mobilenet_v2_140_224/feature_vector/3"
	def feature_extractor(x):
	feature_extractor_module = hub.Module(feature_extractor_url)
	return feature_extractor_module(x)


	IMAGE_SIZE = hub.get_expected_image_size(hub.Module(feature_extractor_url))
	print('Expected size : %s' % IMAGE_SIZE)
	# Expected size : [224, 224]

netsatsawat / CNN_model_architecture.py

Created May 28, 2019 15:43

Part 2 of the script to build the CNN architecture and callbacks function

	np.random.seed(SEED)
	tf.random.set_random_seed(SEED)
	model = Sequential()
	model.add(Conv2D(32, kernel_size=(3, 3), activation='relu', input_shape=(IMG_ROWS, IMG_COLS, 1)))
	model.add(Conv2D(32, (3, 3), activation='relu'))
	model.add(MaxPooling2D(pool_size=(2, 2)))
	model.add(Conv2D(64, (3, 3), activation='relu'))
	model.add(MaxPooling2D(pool_size=(2, 2)))
	model.add(Flatten())
	model.add(Dense(128, activation='relu'))

netsatsawat / CNN_prep.py

Created May 28, 2019 09:41

First part of CNN tutorial on KMNIST data set

	import tensorflow as tf
	from tensorflow.python import keras
	from tensorflow.python.keras.models import Sequential
	from tensorflow.python.keras.layers import Dense, Flatten, Conv2D, Dropout
	from tensorflow.python.keras.layers import MaxPooling2D, BatchNormalization
	from keras.utils.vis_utils import model_to_dot
	from keras.utils import plot_model
	from tensorflow.python.keras.callbacks import ModelCheckpoint
	from tensorflow.keras.callbacks import EarlyStopping
	from sklearn.model_selection import train_test_split

netsatsawat / logistic_regression.py

Created May 19, 2019 20:13

Tune hyperparameters using grid search CV on logit

	cv_params = {'C': [0.001, 0.01, 0.1, 1., 10., 100.],
	'penalty': ['l1', 'l2'],
	'class_weight': [None, 'balanced']
	}
	fix_params = {'random_state': SEED}
	log_cv_1 = GridSearchCV(LogisticRegression(**fix_params), cv_params, scoring='f1', cv=5)
	log_cv_1.fit(X_train, y_train)
	log_clf_all = LogisticRegression({fix_params, **log_cv_1.best_params_})
	_ = myUtilityFunction.prediction_evaluation(log_clf_all, X_train, X_test, y_train, y_test,
	X_train.columns, "coefficients")

netsatsawat / hyperparameter_tuning_xgboost.py

Last active May 19, 2019 18:38

Code snippet to optimize the hyperparameters of XGBoost algorithm

	from sklearn.model_selection import RandomizedSearchCV

	xgb_clf = xgboost.XGBClassifier(random_state=SEED, n_jobs=-1)
	params = {'n_estimators': [50, 100, 200, 300],
	'learning_rate': [0.01, 0.05, 0.1, 0.15],
	'min_child_weight': [1, 2, 3, 5, 10],
	'gamma': [0.1, 0.2, 0.3, 0.4, 0.5, 1],
	'subsample': [0.6, 0.7, 0.8],
	'colsample_bytree': [0.6, 0.7, 0.8],
	'max_depth': [3, 4, 5],

netsatsawat / xgboost_classifier.py

Created May 19, 2019 18:20

Code snippet for xgboost and evaluating the performance

	xgb_clf = xgboost.XGBClassifier(random_state=SEED, n_jobs=-1, learning_rate=0.1,
	max_depth=3, n_estimators=100)
	_ = myUtilityFunction.prediction_evaluation(xgb_clf, X_train, X_test,
	y_train, y_test, X_train.columns, "features")

Satsawat Natakarnkitkul (Net) netsatsawat