def create_net(model, img_prep, img_aug, learning_rate):
if model == "alex":
network = alexNet(img_prep, img_aug, learning_rate)
elif model == "vgg":
network = vggNet(img_prep, img_aug, learning_rate)
elif model == "res":
network = resNet(img_prep, img_aug, learning_rate)
elif model == "alch11":
network = alchNet11(img_prep, img_aug, learning_rate)
elif model == "alch11_without_dropout":
network = alchNet11(img_prep, img_aug, learning_rate, 0)
else:
network = alchNet19(img_prep, img_aug, learning_rate)
return network
def create_net(model, img_prep, img_aug, learning_rate):
if model == "alex":
network = alexNet(img_prep, img_aug, learning_rate)
elif model == "vgg":
network = vggNet(img_prep, img_aug, learning_rate)
elif model == "vggtop5":
network = vggNetTop5(img_prep, img_aug, learning_rate)
elif model == "vgg13":
network = vggNet13(img_prep, img_aug, learning_rate)
elif model == "vgg16":
network = vggNet16(img_prep, img_aug, learning_rate)
elif model == "res":
network = resNet(img_prep, img_aug, learning_rate)
elif model == "alch11":
network = alchNet11(img_prep, img_aug, learning_rate)
elif model == "alch11_without_dropout":
network = alchNet11(img_prep, img_aug, learning_rate, 0)
elif model == "alch_enhance":
network = alchNetEnhance(img_prep, img_aug, learning_rate)
else:
network = alchNet19(img_prep, img_aug, learning_rate)
return network
batch_size=batch_size,
n_channels=1,
n_classes=n_classes,
shuffle=True)
validation_generator = volume.DataGenerator(validationlistIDs,
validationdataDir,
v_size=sideLength,
batch_size=batch_size,
n_channels=1,
n_classes=n_classes,
shuffle=True)
##------------------------------ Model ---------------------------------------##
#Create
model = models.alexNet(sideLength)
model.summary()
#Track accuracy and loss in real-time
#if jupyter notebook:
log_dir = "/media/data_crypt_2/alexNet" + datetime.datetime.now().strftime(
"%Y-%m-%d_%H:%M:%S")
file_writer = tf.summary.create_file_writer(log_dir + "/metrics")
file_writer.set_as_default()
#python script
saving_path = log_dir
if not os.path.isdir(saving_path):
os.mkdir(saving_path)
saving_path = os.path.join(saving_path, 'metricsHistory')
if not os.path.isdir(saving_path):
]
epochs = [i for i in range(1, 51)] #CHANGE THIS
a = len(testinglistIDs)
print('total of {} testing images'.format(a))
step = math.ceil(a / 16)
batch_size = 128 #CHANGE THIS
sideLength = 96 #CHANGE THIS
test_generator = testDataGenerator(testinglistIDs,
testDir,
batch_size=batch_size,
v_size=sideLength)
model = models.alexNet(sideLength) #CHANGE THIS
for epoch in epochs:
model.load_weights(
"/data/resized_FPR/models/improved_res_alex_aug/2021-02-16_11:57:24/checkpoints/alex_aug_{}.hd5f"
.format(str(epoch).zfill(2))) #CHANGE THIS
opt = SGD()
model.compile(optimizer=opt,
loss='binary_crossentropy',
metrics=['accuracy']) #CHANGE THIS (OPTIMIZER)
prediction = model.evaluate(test_generator, verbose=1)
savePath = '/data/resized_FPR/models/improved_res_alex_aug/2021-02-16_11:57:24/evaluation' #CHANGE THIS
if not os.path.isdir(savePath): os.mkdir(savePath)
with open(os.path.join(savePath, 'eval_epoch{}.txt'.format(epoch)),
'w+') as f:
if model == 'res18conv':
# ResNet18 Conv
model, criterion, optimizer, scheduler = models.resNet18_conv()
model_conv = fit(model, criterion, optimizer, scheduler, num_epochs=30)
predict(model_conv)
if model == 'res152conv':
# ResNet152 Conv
model, criterion, optimizer, scheduler = models.resNet152_conv()
model_conv = fit(model, criterion, optimizer, scheduler, num_epochs=30)
predict(model_conv)
if model == 'dense161':
# DenseNet161
model, criterion, optimizer, scheduler = models.denseNet161_ft()
model_ft = fit(model, criterion, optimizer, scheduler, num_epochs=30)
predict(model_ft)
if model == 'vgg19':
# VGG 19-layer model
model, criterion, optimizer, scheduler = models.vgg19()
model_ft = fit(model, criterion, optimizer, scheduler, num_epochs=30)
predict(model_ft)
if model == 'alex':
# AlexNet
model, criterion, optimizer, scheduler = models.alexNet()
model_ft = fit(model, criterion, optimizer, scheduler, num_epochs=30)
predict(model_ft)
