def test_resnet50_caffe(self):
if self.data_dir is None:
unittest.TestCase.skipTest(self, "DLPY_DATA_DIR is not set in the environment variables")
file_dependency = self.data_dir + 'ResNet-50-model.caffemodel.h5'
if not file_exist_on_server(self.s, file_dependency):
unittest.TestCase.skipTest(self, "File, {}, not found.".format(file_dependency))
model = ResNet50_Caffe(self.s, n_channels=3, height=224, random_flip='HV',
pre_trained_weights_file=self.data_dir + 'ResNet-50-model.caffemodel.h5',
pre_trained_weights=True,
include_top=False,
n_classes=120,
random_crop='unique')
model.print_summary()
model = ResNet50_Caffe(self.s, n_channels=3, height=224, random_flip='HV',
pre_trained_weights_file=self.data_dir + 'ResNet-50-model.caffemodel.h5',
pre_trained_weights=True,
include_top=False,
n_classes=120,
random_crop=None,
offsets=None)
model.print_summary()
# test random_mutation and crop on VDMML 8.4
model = ResNet50_Caffe(self.s, n_channels=3, height=224, random_flip='HV',
pre_trained_weights_file=self.data_dir + 'ResNet-50-model.caffemodel.h5',
pre_trained_weights=True,
include_top=False,
n_classes=120,
random_crop='RESIZETHENCROP',
random_mutation='random',
offsets=None)
def test_resnet50_3(self):
from dlpy.applications import ResNet50_Caffe
reshape = Reshape(width=224, height=224, depth=3, order='WHD')
model = ResNet50_Caffe(self.s, reshape_after_input=reshape)
model.print_summary()
# test it with pretrained weights
model1 = ResNet50_Caffe(self.s,
model_table='Resnet50',
n_classes=1000,
n_channels=3,
width=224,
height=224,
scale=1,
offsets=None,
random_crop='unique',
random_flip='hv',
random_mutation='random',
pre_trained_weights=True,
pre_trained_weights_file=self.data_dir +
'ResNet-50-model.caffemodel.h5',
include_top=True,
reshape_after_input=reshape)
res = model1.print_summary()
print(res)
self.assertEqual(res.iloc[1, 6][0], 224)
self.assertEqual(res.iloc[1, 6][1], 224)
self.assertEqual(res.iloc[1, 6][2], 3)
def test_resnet50_caffe_caslib_msg(self):
if self.data_dir is None:
unittest.TestCase.skipTest(
self, "DLPY_DATA_DIR is not set in the environment variables")
model = ResNet50_Caffe(self.s,
n_channels=3,
height=224,
random_flip='HV',
pre_trained_weights_file=self.data_dir +
'data/ResNet-50-model.caffemodel.h5',
pre_trained_weights=True,
include_top=False,
n_classes=120,
random_crop='unique')
model = ResNet50_Caffe(self.s,
n_channels=3,
height=224,
random_flip='HV',
pre_trained_weights_file=self.data_dir +
'ResNet-50-model.caffemodel.h5',
pre_trained_weights=True,
include_top=False,
n_classes=120,
random_crop='unique')
model.print_summary()
def test_resnet50_3(self):
from dlpy.applications import ResNet50_Caffe
if self.data_dir is None:
unittest.TestCase.skipTest(self, "DLPY_DATA_DIR is not set in the environment variables")
file_dependency = self.data_dir + 'ResNet-50-model.caffemodel.h5'
if not file_exist_on_server(self.s, file_dependency):
unittest.TestCase.skipTest(self, "File, {}, not found.".format(file_dependency))
reshape = Reshape(width=224, height=224, depth=3, order='WHD')
model = ResNet50_Caffe(self.s, reshape_after_input=reshape)
model.print_summary()
# test it with pretrained weights
model1 = ResNet50_Caffe(self.s, model_table='Resnet50', n_classes=1000, n_channels=3,
width=224, height=224, scale=1,
offsets=None,
random_crop='unique',
random_flip='hv',
random_mutation='random',
pre_trained_weights=True,
pre_trained_weights_file=self.data_dir + 'ResNet-50-model.caffemodel.h5',
include_top=True,
reshape_after_input=reshape)
res = model1.print_summary()
print(res)
self.assertEqual(res.iloc[1, 6][0], 224)
self.assertEqual(res.iloc[1, 6][1], 224)
self.assertEqual(res.iloc[1, 6][2], 3)
def test_resnet50_4(self):
from dlpy.applications import ResNet50_Caffe
reshape = Pooling(width=2, height=2, stride=2)
self.assertRaises(
DLPyError,
lambda: ResNet50_Caffe(self.s, reshape_after_input=reshape))
def test_resnet50_layerid(self):
if self.data_dir is None:
unittest.TestCase.skipTest(self, "DLPY_DATA_DIR is not set in the environment variables")
model = ResNet50_Caffe(self.s)
model.print_summary()
model.print_summary()
def test_mix_cnn_rnn_network(self):
from dlpy.applications import ResNet50_Caffe
from dlpy import Sequential
from dlpy.blocks import Bidirectional
# the case is to test if CNN and RNN model can be connect using functional api
# the model_type is expected to be RNN in 19w47.
# CNN
model = ResNet50_Caffe(self.s)
cnn_head = model.to_functional_model(stop_layers = model.layers[-1])
# RNN
model_rnn = Sequential(conn = self.s, model_table = 'rnn')
model_rnn.add(Bidirectional(n = 100, n_blocks = 2))
model_rnn.add(OutputLayer('fixed'))
f_rnn = model_rnn.to_functional_model()
# connecting
inp = Input(**cnn_head.layers[0].config)
x = cnn_head(inp)
y = f_rnn(x)
cnn_rnn = Model(self.s, inp, y)
cnn_rnn.compile()
# check type
self.assertTrue(cnn_rnn.model_type == 'RNN')
self.assertTrue(cnn_rnn.layers[-1].name == 'fixed')
self.assertTrue(x[0].shape == (1, 1, 2048))
f_rnn = model_rnn.to_functional_model()
# connecting
inp = Input(**cnn_head.layers[0].config)
x = cnn_head(inp)
y = f_rnn(x)
cnn_rnn = Model(self.s, inp, y)
cnn_rnn.compile()
# it should be fixed if I create f_rnn again.
self.assertTrue(cnn_rnn.layers[-1].name == 'fixed')
inp = Input(**cnn_head.layers[0].config)
x = cnn_head(inp)
y = f_rnn(x)
cnn_rnn = Model(self.s, inp, y)
cnn_rnn.compile()
# it should be fixed if I create f_rnn again.
self.assertTrue(cnn_rnn.layers[-1].name == 'fixed_2')
def test_heat_map_analysis(self):
if self.data_dir is None:
unittest.TestCase.skipTest(self, 'DLPY_DATA_DIR is not set in the environment variables')
if not file_exist_on_server(self.s, self.data_dir + 'ResNet-50-model.caffemodel.h5'):
unittest.TestCase.skipTest(self, "File, {}, not found.".format(self.data_dir
+ 'ResNet-50-model.caffemodel.h5'))
from dlpy.applications import ResNet50_Caffe
from dlpy.images import ImageTable
pre_train_weight_file = os.path.join(self.data_dir, 'ResNet-50-model.caffemodel.h5')
my_im = ImageTable.load_files(self.s, self.data_dir+'giraffe_dolphin_small')
my_im_r = my_im.resize(width=224, inplace=False)
model = ResNet50_Caffe(self.s, model_table='ResNet50_Caffe',
n_classes=2, n_channels=3, width=224, height=224, scale=1,
random_flip='none', random_crop='none',
offsets=my_im_r.channel_means, pre_trained_weights=True,
pre_trained_weights_file=pre_train_weight_file,
include_top=False)
model.fit(data=my_im_r, mini_batch_size=1, max_epochs=1)
model.heat_map_analysis(data=my_im_r, mask_width=None, mask_height=None, step_size=None,
max_display=1)
self.assertRaises(ValueError, lambda:model.heat_map_analysis(mask_width=56, mask_height=56,
step_size=8, display=False))
self.assertRaises(ValueError, lambda:model.heat_map_analysis(data=my_im, mask_width=56,
mask_height=56, step_size=8, display=False))
try:
from numpy import array
except:
unittest.TestCase.skipTest(self, 'numpy is not installed')
self.assertRaises(ValueError, lambda:model.heat_map_analysis(data=array([]), mask_width=56,
mask_height=56, step_size=8, display=False))
def test_resnet50_2(self):
from dlpy.applications import ResNet50_Caffe
model = ResNet50_Caffe(self.s)
model.print_summary()
def test_multiple_stop_layers1(self):
from dlpy.applications import ResNet50_Caffe
resnet50 = ResNet50_Caffe(self.s, "res50")
stop_layers = [resnet50.layers[x] for x in [-2, 4, -8] ]
feature_extractor1 = resnet50.to_functional_model(stop_layers=stop_layers)
