def test_add_output_layer():
model1 = dpp.ClassificationModel()
model2 = dpp.SemanticSegmentationModel()
model3 = dpp.CountCeptionModel()
model1.set_image_dimensions(5, 5, 3)
model2.set_image_dimensions(5, 5, 3)
with pytest.raises(RuntimeError):
model1.add_output_layer(2.5, 3)
model1.add_input_layer()
model2.add_input_layer()
model3.add_input_layer()
with pytest.raises(TypeError):
model1.add_output_layer("2")
with pytest.raises(ValueError):
model1.add_output_layer(-0.4)
with pytest.raises(TypeError):
model1.add_output_layer(2.0, 3.4)
with pytest.raises(ValueError):
model1.add_output_layer(2.0, -4)
with pytest.raises(RuntimeError):
model2.add_output_layer(
output_size=3
) # Semantic segmentation needed for this runtime error to occur
model1.add_output_layer(2.5, 3)
assert isinstance(model1._last_layer(), dpp.layers.fullyConnectedLayer)
with pytest.warns(Warning):
model2.add_output_layer(regularization_coefficient=2.0)
assert isinstance(model2._last_layer(), dpp.layers.convLayer)
model3.add_output_layer()
assert isinstance(model3._last_layer(), dpp.layers.inputLayer)
def __init__(self, batch_size=1, image_height=300, image_width=300, image_depth=3):
"""A network which counts flowers in plant images"""
m_path, _ = os.path.split(__file__)
checkpoint_path = os.path.join(m_path, 'network_states', self.__dir_name)
import deepplantphenomics as dpp
self.model = dpp.CountCeptionModel(debug=True, load_from_saved=checkpoint_path)
# Define model hyperparameters
self.model.set_loss_function('l1')
self.model.set_batch_size(batch_size)
self.model.set_number_of_threads(4)
self.model.set_image_dimensions(image_height, image_width, image_depth)
# Define a model architecture
self.model.add_input_layer()
self.model.add_convolutional_layer(filter_dimension=[3, 3, 3, 64],
stride_length=1,
activation_function='lrelu',
padding=self.patch_size,
batch_norm=True,
epsilon=1e-5,
decay=0.9)
self.model.add_paral_conv_block(filter_dimension_1=[1, 1, 0, 16],
filter_dimension_2=[3, 3, 0, 16])
self.model.add_paral_conv_block(filter_dimension_1=[1, 1, 0, 16],
filter_dimension_2=[3, 3, 0, 32])
self.model.add_convolutional_layer(filter_dimension=[14, 14, 0, 16],
stride_length=1,
activation_function='lrelu',
padding=0,
batch_norm=True,
epsilon=1e-5,
decay=0.9)
self.model.add_paral_conv_block(filter_dimension_1=[1, 1, 0, 112],
filter_dimension_2=[3, 3, 0, 48])
self.model.add_paral_conv_block(filter_dimension_1=[1, 1, 0, 64],
filter_dimension_2=[3, 3, 0, 32])
self.model.add_paral_conv_block(filter_dimension_1=[1, 1, 0, 40],
filter_dimension_2=[3, 3, 0, 40])
self.model.add_paral_conv_block(filter_dimension_1=[1, 1, 0, 32],
filter_dimension_2=[3, 3, 0, 96])
self.model.add_convolutional_layer(filter_dimension=[18, 18, 0, 32],
stride_length=1,
activation_function='lrelu',
padding=0,
batch_norm=True,
epsilon=1e-5,
decay=0.9)
self.model.add_convolutional_layer(filter_dimension=[1, 1, 0, 64],
stride_length=1,
activation_function='lrelu',
padding=0,
batch_norm=True,
epsilon=1e-5,
decay=0.9)
self.model.add_convolutional_layer(filter_dimension=[1, 1, 0, 64],
stride_length=1,
activation_function='lrelu',
padding=0,
batch_norm=True,
epsilon=1e-5,
decay=0.9)
self.model.add_convolutional_layer(filter_dimension=[1, 1, 0, 1],
stride_length=1,
activation_function='lrelu',
padding=0,
batch_norm=True,
epsilon=1e-5,
decay=0.9)
model.set_patch_size(1.0, 1)
with pytest.raises(ValueError):
model.set_patch_size(-1, 1)
with pytest.raises(TypeError):
model.set_patch_size(1, 1.0)
with pytest.raises(ValueError):
model.set_patch_size(1, -1)
@pytest.mark.parametrize(
"model,bad_loss,good_loss",
[(dpp.ClassificationModel(), 'l2', 'softmax cross entropy'),
(dpp.RegressionModel(), 'softmax cross entropy', 'l2'),
(dpp.SemanticSegmentationModel(), 'l2', 'sigmoid cross entropy'),
(dpp.ObjectDetectionModel(), 'l2', 'yolo'),
(dpp.CountCeptionModel(), 'l2', 'l1'),
(dpp.HeatmapObjectCountingModel(), 'l1', 'sigmoid cross entropy')])
def test_set_loss_function(model, bad_loss, good_loss):
with pytest.raises(TypeError):
model.set_loss_function(0)
with pytest.raises(ValueError):
model.set_loss_function(bad_loss)
model.set_loss_function(good_loss)
def test_set_yolo_parameters():
model = dpp.ObjectDetectionModel()
with pytest.raises(RuntimeError):
model.set_yolo_parameters()
model.set_image_dimensions(448, 448, 3)
model.set_yolo_parameters()
#
# Used to train the Count-ception model
#
import deepplantphenomics as dpp
model = dpp.CountCeptionModel(debug=True,
load_from_saved=False,
save_checkpoints=False,
report_rate=20)
patch_size = 32
# Setup and hyperparameters
model.set_loss_function('l1')
model.set_batch_size(2)
model.set_number_of_threads(4)
model.set_image_dimensions(300, 300, 3)
model.set_test_split(0.3)
model.set_validation_split(0.2)
model.set_learning_rate(0.0001)
model.set_weight_initializer('xavier')
model.set_maximum_training_epochs(10)
# Load images and ground truth from a pickle file
model.load_countception_dataset_from_pkl_file('MBM-dataset.pkl')
# Define a model architecture
model.use_predefined_model("countception")