def test_resnet3d_152(resnet3d_test):
"""Test 152."""
K.set_image_data_format('channels_last')
model = Resnet3DBuilder.build_resnet_152((224, 224, 224, 1), 2)
resnet3d_test(model)
K.set_image_data_format('channels_first')
model = Resnet3DBuilder.build_resnet_152((1, 512, 512, 256), 2)
resnet3d_test(model)
def test_get_block():
"""Test get residual block."""
K.set_image_data_format('channels_last')
Resnet3DBuilder.build((224, 224, 224, 1), 2, 'bottleneck',
[2, 2, 2, 2], reg_factor=1e-4)
assert True
with pytest.raises(ValueError):
Resnet3DBuilder.build((224, 224, 224, 1), 2, 'nullblock',
[2, 2, 2, 2], reg_factor=1e-4)
def test_resnet3d_18():
K.set_image_data_format('channels_last')
model = Resnet3DBuilder.build_resnet_18((224, 224, 224, 1), 2)
model.compile(loss="categorical_crossentropy", optimizer="sgd")
assert True, "Failed to build with tensorflow"
K.set_image_data_format('channels_first')
model = Resnet3DBuilder.build_resnet_18((1, 512, 512, 256), 2)
model.compile(loss="categorical_crossentropy", optimizer="sgd")
assert True, "Failed to build with theano"
def create_reg_resnet18_3D(img_x, img_y, ch_n, seq_len, tgt_size):
resnet18_3D = Resnet3DBuilder.build_resnet_18(
(seq_len, img_x, img_y, ch_n), tgt_size)
resnet18_3D.layers.pop()
layer = Dense(32, activation='relu')(resnet18_3D.output)
resnet18_3D.layers[-1].outbound_nodes = []
resnet18_3D.outputs = [resnet18_3D.layers[-1].output]
output = resnet18_3D.get_layer('flatten_1').output
output = Dense(32, activation='relu')(output)
output = Dense(1, activation='linear')(output)
reg_resnet18_3D = Model(resnet18_3D.input, output)
return reg_resnet18_3D
batch_x.append(frames)
batch_y.append(labels)
batch_x = np.array(batch_x)
batch_y = np.array(batch_y)
batch_x = batch_x.reshape(-1, 30, 64, 96, 3)
yield batch_x, batch_y
train_generator = generator('train', batch_size=64)
validation_generator = generator('val', batch_size=64)
model = Resnet3DBuilder.build_resnet_101((30, 64, 96, 3), 27)
model.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy'])
callbacks = [
ModelCheckpoint('./resnet3d_50_dropout_3c_64x96_b32.h5',
monitor='val_accuracy',
verbose=1,
save_best_only=True),
EarlyStopping(patience=5)
]
history = model.fit(train_generator,
validation_data=validation_generator,
steps_per_epoch=119_000 // 64,
validation_steps=15_000 // 64,
callbacks=callbacks,
**init_args['volume_image_data_generator']['val']['init'])
train_vol_loader = NPYDataLoader(
**init_args['volume_image_data_loader']['train'])
val_vol_loader = NPYDataLoader(**init_args['volume_image_data_loader']['val'])
train_iter_args = init_args['volume_image_data_generator']['train'][
'flow_from_loader']
train_iter_args['volume_image_data_loader'] = train_vol_loader
val_iter_args = init_args['volume_image_data_generator']['val'][
'flow_from_loader']
val_iter_args['volume_image_data_loader'] = val_vol_loader
image_shape = train_datagen.image_shape
regularization_factor = 1
model = Resnet3DBuilder.build_resnet_18(image_shape, nb_classes,
regularization_factor)
model.load_weights(weights)
compile_args = init_args['model']['compile']
compile_args['optimizer'] = Adam(lr=1e-4)
model.compile(**compile_args)
model_fit_args = init_args['model']['fit_generator']
model_fit_args['generator'] = train_datagen.flow_from_loader(**train_iter_args)
model_fit_args['validation_data'] = val_datagen.flow_from_loader(
**val_iter_args)
model_fit_args['callbacks'] = [
checkpointer, lr_reducer, early_stopper, csv_logger
]
model.fit_generator(**model_fit_args)
model.save('output/resnet18_{}_ctd.h5'.format(title))
print("Valid hdf5 file in 'read' mode: " + str(hdf5_file))
file_size = os.path.getsize(path_to_hdf5)
print('Size of hdf5 file: {:.3f} GB'.format(file_size/2.0**30))
num_rows = hdf5_file['input'].shape[0]
print("There are {} images in the dataset.".format(num_rows))
print("The datasets within the HDF5 file are:\n {}".format(list(hdf5_file.values())))
input_shape = tuple(list(hdf5_file["input"].attrs["lshape"]))
batch_size = args.batchsize # Batch size to use
print ("Input shape of tensor = {}".format(input_shape))
from resnet3d import Resnet3DBuilder
model = Resnet3DBuilder.build_resnet_18((64, 64, 64, 1), 1) # (input tensor shape, number of outputs)
tb_log = keras.callbacks.TensorBoard(log_dir=TB_LOG_DIR,
histogram_freq=0,
batch_size=batch_size,
write_graph=True,
write_grads=True,
write_images=True,
embeddings_freq=0,
embeddings_layer_names=None,
embeddings_metadata=None)
checkpointer = keras.callbacks.ModelCheckpoint(filepath=CHECKPOINT_FILENAME,
monitor="val_loss",
verbose=1,
copy(__file__, output_dir)
# Download the data from: /hpf/largeprojects/ccm/devin/plastics-data/???
datapath = "data/data.h5"
test_size = 0.20
X_train, X_test, Y_train, Y_test = create_train_test(datapath, test_size)
print ("number of training examples = " + str(X_train.shape[0]))
print ("number of test examples = " + str(X_test.shape[0]))
print ("X_train shape: " + str(X_train.shape))
print ("Y_train shape: " + str(Y_train.shape))
print ("X_test shape: " + str(X_test.shape))
print ("Y_test shape: " + str(Y_test.shape))
model = Resnet3DBuilder.build_resnet_50((128, 128, 128, 1), 3)
# model = Resnet3DBuilder.build((128, 128, 128, 1), 3, basic_block, [1, 1, 1, 1], reg_factor=1e-4)
# Uncomment the following block to do transfer learning!
model.layers[-1].name = "dense_resnet_1" # Rename final dense layer so correct number of output classes is used (3 instead of MNIST 10)
# MNIST build_resnet_18
# model.load_weights('/home/carsonmclean/dev/csc2541/csc2541/output/04-03-03:37/model.h5',
# by_name = True)
# MNIST [1,1,1,1]
# model.load_weights('/home/carsonmclean/dev/csc2541/csc2541/output/04-09-00:25/model.h5',
# by_name = True)
# ModelNet40/PointNet build_resnet_18 dim = 64
# model.load_weights('/home/carsonmclean/dev/csc2541/csc2541/output/04-09-17:40/model.h5',
# by_name = True)
# ModelNet40/PointNet build_resnet_18 dim = 128
#print (len(y_train))
del labels
# batch_size =1
x_val = np.array(x_val)
#print ("byeeeee",x_train.shape[0])
y_val = np.array(y_val)
print ('=++=++=++=++=',y_val)
x_val,y_val = shuffle(x_val,y_val)
#print ("////////",x_val)
x_val = x_val.astype("float32")
y_val = y_val.astype("float32")
x_val /= 255
#y_val /= 255
yield x_val, y_val
x_val = []
y_val = []
from resnet3d import Resnet3DBuilder
model = Resnet3DBuilder.build_resnet_10((3, 224, 224, 3), 101)
model.summary()
#optimizer = keras.optimizers.RMSprop(lr=0.001, rho=0.9, epsilon=None, decay=0.0)
optimizer = keras.optimizers.Adam(lr=0.0001, beta_1=0.9, beta_2=0.999, epsilon=None, decay=0.0, amsgrad=False)
model.compile(optimizer=optimizer,
loss='sparse_categorical_crossentropy',
metrics=['categorical_accuracy'])
#model.fit_generator(dataGenerator(train_lines),steps_per_epoch=7058, epochs=2)
model.fit_generator(generator=dataGenerator(train_lines), validation_data=val_dataGenerator(val_lines), validation_steps=2472, steps_per_epoch=7058, epochs=5)
y_filename = os.path.join(y_path, filename)
X = np.load(X_fileanme)
y = np.load(y_filename)
for i in range(X.shape[0] // batch_size):
indexes = np.random.randint(X.shape[0], size=batch_size)
yield X[indexes], y[indexes]
del X
del y
if __name__ == '__main__':
# model, training_model = get_unet_model()
# model, training_model = get_model()
model = Resnet3DBuilder.build_resnet_50((64, 64, 16, 1), 2)
training_model = keras.utils.multi_gpu_model(model)
training_model.compile(optimizer=Adam(amsgrad=True), loss='binary_crossentropy', metrics=['accuracy'])
model.summary()
callbacks=[
RedirectModel(keras.callbacks.ModelCheckpoint(
os.path.join(
'./model_checkpoints',
'{epoch:02d}.h5'
),
verbose=1,
), model),
keras.callbacks.TensorBoard(
log_dir='./logs/' + datetime.datetime.now().strftime('%Y%m%d%H%M')
),
import numpy as np from resnet3d import Resnet3DBuilder # pseudo volumetric data X_train = np.random.rand(10, 64, 64, 32, 1) labels = np.random.randint(0, 2, size=[10]) y_train = np.eye(2)[labels] # train model = Resnet3DBuilder.build_resnet_18((64, 64, 32, 1), 2) model.compile(loss="categorical_crossentropy", optimizer="sgd") model.fit(X_train, y_train, batch_size=10)
## convert to 1 + 4D space (1st argument represents number of rows in the dataset)
X_train = X_train.reshape(X_train.shape[0], 16, 16, 16, 1)
X_test = X_test.reshape(X_test.shape[0], 16, 16, 16, 1)
Y_train = to_categorical(Y_train, 10)
Y_test = to_categorical(Y_test, 10)
print("number of training examples = " + str(X_train.shape[0]))
print("number of test examples = " + str(X_test.shape[0]))
print("X_train shape: " + str(X_train.shape))
print("Y_train shape: " + str(Y_train.shape))
print("X_test shape: " + str(X_test.shape))
print("Y_test shape: " + str(Y_test.shape))
# model = Resnet3DBuilder.build_resnet_18((16, 16, 16, 1), 10)
model = Resnet3DBuilder.build((16, 16, 16, 1),
10,
basic_block, [1, 1, 1, 1],
reg_factor=1e-4)
adam = Adam(lr=0.0001)
model.compile(optimizer=adam,
loss='categorical_crossentropy',
metrics=['accuracy'])
earlystop = EarlyStopping(monitor='val_acc',
min_delta=0.001,
patience=5,
verbose=2,
restore_best_weights=True)
callback_list = [earlystop]
model_info = model.fit(X_train,
Y_train,
epochs=100,
import numpy as np from resnet3d import Resnet3DBuilder # pseudo volumetric data X_train = np.random.rand(10, 64, 64, 32, 1) labels = np.random.randint(0, 2, size=[10]) y_train = np.eye(2)[labels] # train model = Resnet3DBuilder.build_resnet_50((64, 64, 32, 1), 2, multilabel=True) model.compile(loss="categorical_crossentropy", optimizer="sgd") model.fit(X_train, y_train, batch_size=10)
copy(__file__, output_dir)
datapath = "data/modelnet/modelnet40_ply_hdf5_2048/"
test_size = 0.25
X_train, X_test, Y_train, Y_test = create_train_test(datapath, test_size)
print ("number of training examples = " + str(X_train.shape[0]))
print ("number of test examples = " + str(X_test.shape[0]))
print ("X_train shape: " + str(X_train.shape))
print ("Y_train shape: " + str(Y_train.shape))
print ("X_test shape: " + str(X_test.shape))
print ("Y_test shape: " + str(Y_test.shape))
dim = 64
model = Resnet3DBuilder.build_resnet_50((dim, dim, dim, 1), 40)
# model = Resnet3DBuilder.build((128, 128, 128, 1), 3, basic_block, [1, 1, 1, 1], reg_factor=1e-4)
# Uncomment the following block to do transfer learning!
# model.layers[-1].name = "dense_resnet_1" # Rename final dense layer so correct number of output classes is used (3 instead of MNIST 10)
# model.load_weights('/home/carsonmclean/dev/csc2541/csc2541/output/04-03-03:37/model.h5',
# by_name = True) # build_resnet_18
# model.load_weights('/home/carsonmclean/dev/csc2541/csc2541/output/04-09-00:25/model.h5',
# by_name = True) # [1,1,1,1]
# End of transfer learning
adam = Adam(lr=0.00001)
model.compile(optimizer=adam,
loss='categorical_crossentropy',
metrics=['accuracy'])
earlystop = EarlyStopping(monitor = 'val_acc',
def test_bad_shape():
"""Input shape need to be 4."""
K.set_image_data_format('channels_last')
with pytest.raises(ValueError):
Resnet3DBuilder.build_resnet_152((224, 224, 224), 2)
def model_confirm(self, choosed_model):
if choosed_model == 'VGG16':
model = MODEL(self.config).VGG16()
elif choosed_model == 'VGG19':
model = MODEL(self.config).VGG19()
elif choosed_model == 'ResNet3D':
model = Resnet3DBuilder().build_resnet_18((91, 109, 91, 1), 2)
elif choosed_model == 'AlexNet':
model = MODEL(self.config).AlexNet()
elif choosed_model == 'LeNet':
model = MODEL(self.config).LeNet()
elif choosed_model == 'ZF_Net':
model = MODEL(self.config).ZF_Net()
elif choosed_model == 'ResNet18':
model = ResnetBuilder().build_resnet18(self.config)
elif choosed_model == 'ResNet34':
model = ResnetBuilder().build_resnet34(self.config)
elif choosed_model == 'ResNet101':
model = ResnetBuilder().build_resnet101(self.config)
elif choosed_model == 'ResNet152':
model = ResnetBuilder().build_resnet152(self.config)
elif choosed_model == 'mnist_net':
model = MODEL(self.config).mnist_net()
elif choosed_model == 'TSL16':
model = MODEL(self.config).TSL16()
elif choosed_model == 'ResNet50':
model = keras.applications.ResNet50(include_top=True,
weights=None,
input_tensor=None,
input_shape=(self.normal_size,
self.normal_size,
self.channles),
pooling='max',
classes=self.classNumber)
elif choosed_model == 'InceptionV3':
model = keras.applications.InceptionV3(
include_top=True,
weights=None,
input_tensor=None,
input_shape=(self.normal_size, self.normal_size,
self.channles),
pooling='max',
classes=self.classNumber)
elif choosed_model == 'Xception':
model = keras.applications.Xception(include_top=True,
weights=None,
input_tensor=None,
input_shape=(self.normal_size,
self.normal_size,
self.channles),
pooling='max',
classes=self.classNumber)
elif choosed_model == 'MobileNet':
model = keras.applications.MobileNet(include_top=True,
weights=None,
input_tensor=None,
input_shape=(self.normal_size,
self.normal_size,
self.channles),
pooling='max',
classes=self.classNumber)
# elif choosed_model == 'DenseNet':
# depth = 40
# nb_dense_block = 3
# growth_rate = 12
# nb_filter = 12
# bottleneck = False
# reduction = 0.0
# dropout_rate = 0.0
#
# img_dim = (self.channles, self.normal_size) if K.image_dim_ordering() == "th" else (
# self.normal_size, self.normal_size, self.channles)
#
# model = densenet.DenseNet(img_dim, classNumber=self.classNumber, depth=depth, nb_dense_block=nb_dense_block,
# growth_rate=growth_rate, nb_filter=nb_filter, dropout_rate=dropout_rate,
# bottleneck=bottleneck, reduction=reduction, weights=None)
return model
