def make_model(network, input_shape):
if network == 'resnet101_softmax':
return resnet101_fpn(input_shape, channels=3, activation="softmax")
elif network == 'resnet152_2':
return resnet152_fpn(input_shape, channels=2, activation="sigmoid")
elif network == 'resnet101_2':
return resnet101_fpn(input_shape, channels=2, activation="sigmoid")
elif network == 'resnet50_2':
return resnet50_fpn(input_shape, channels=2, activation="sigmoid")
elif network == 'resnetv2':
return inception_resnet_v2_fpn(input_shape,
channels=2,
activation="sigmoid")
elif network == 'resnetv2_3':
return inception_resnet_v2_fpn(input_shape,
channels=3,
activation="sigmoid")
elif network == 'densenet169':
return densenet_fpn(input_shape, channels=2, activation="sigmoid")
elif network == 'densenet169_softmax':
return densenet_fpn(input_shape, channels=3, activation="softmax")
elif network == 'resnet101_unet_2':
return resnet101_fpn(input_shape, channels=2, activation="sigmoid")
elif network == 'xception_fpn':
return xception_fpn(input_shape, channels=2, activation="sigmoid")
elif network == 'resnet50_2':
return resnet50_fpn(input_shape, channels=2, activation="sigmoid")
else:
raise ValueError('unknown network ' + network)
def make_model(network, black_detect=False):
input_shape = (None, None, 3)
if network == "xception_fpn":
return xception_fpn(input_shape,
channels=1,
activation="sigmoid",
black_detect=black_detect)
elif network == "resnet152":
return resnet152_fpn(input_shape,
channels=1,
activation="sigmoid",
black_detect=black_detect)
elif network == "inception_resnet_v2":
return inception_resnet_v2_fpn(input_shape,
channels=1,
activation="sigmoid",
black_detect=black_detect)
elif network == "densenet169":
return densenet_fpn(input_shape,
channels=1,
activation="sigmoid",
black_detect=black_detect)
elif network == "testnet":
return testnet(input_shape,
channels=1,
activation="sigmoid",
black_detect=black_detect)
else:
raise NotImplementedError("Network {} not implement.".format(network))