def test_train(self):
resnet = ResNet(dataset=DummyDataset(batch_size=2),
block_nums=1,
epochs=1)
history = resnet.train()
ok_('loss' in history)
def test_train(self):
fcnn = FCNNClassifier(dataset=DummyDataset(batch_size=2),
hidden_nums=16,
dropout_rate=0.8)
history = fcnn.train()
ok_('loss' in history)
def test_load_pretrained_model(self):
path = pathlib.Path(__file__).parent.joinpath('test_path')
path.mkdir(parents=True, exist_ok=True)
yolo = YoloV2(dataset=DummyDataset(), classification=True, epochs=1)
history = yolo.train()
yolo.save(path.joinpath('model'))
weight = yolo.model.layers[2].weights[0]
yolo = YoloV2(dataset=OjbjectDetectionDummyDataset(),
restore_path=path.joinpath('model'))
ok_((np.abs(weight - yolo.model.layers[2].weights[0]) <= 1e-5).all())
shutil.rmtree(path)
def test_init(self):
yolo = YoloV2(dataset=OjbjectDetectionDummyDataset())
eq_(yolo.model.outputs[0].shape.as_list(), [None, 13, 13, 5 * (2 + 5)])
yolo = YoloV2(dataset=OjbjectDetectionDummyDataset(), tiny=False)
eq_(yolo.model.outputs[0].shape.as_list(), [None, 13, 13, 5 * (2 + 5)])
yolo = YoloV2(dataset=DummyDataset(), classification=True)
eq_(yolo.model.outputs[0].shape.as_list(),
[None, DummyDataset.category_nums])
yolo = YoloV2(dataset=OjbjectDetectionDummyDataset(),
tiny=False,
classification=True)
eq_(yolo.model.outputs[0].shape.as_list(),
[None, DummyDataset.category_nums])
def test_init(self):
resnet = ResNet(dataset=DummyDataset(), block_nums=1)
eq_(len(resnet.blocks), 3)
def test_init_usext(self):
resnet = ResNet(dataset=DummyDataset(), block_nums=1, use_xt=True)
ok_(isinstance(resnet.blocks[0]['residual_path'][-4], list))
def test_init_usese(self):
resnet = ResNet(dataset=DummyDataset(), block_nums=1, use_se=True)
ok_(isinstance(resnet.blocks[0]['residual_path'][-1], SEBlock))
def test_init(self):
efficientnet = EfficientNet(dataset=DummyDataset(), block_nums=1)
def test_train(self):
efficientnet = EfficientNet(dataset=DummyDataset(), block_nums=1)
history = efficientnet.train()
ok_('loss' in history)
def test_train_classification(self):
yolo = YoloV2(dataset=DummyDataset(), classification=True)
history = yolo.train()
ok_('loss' in history)
def test_init(self):
fcnn = FCNNClassifier(dataset=DummyDataset(),
hidden_nums=16,
dropout_rate=0.8)
eq_(fcnn.dense1.kernel.shape[1], 16)
eq_(fcnn.dropout.rate, 0.8)