def evaluate(auto: bool = False):
"""Evaluation function"""
logger = logging.getLogger()
logger.setLevel(logging.INFO)
print("Loading images... ", end="")
# Load images
images = load_images(range(1, 15))
print("OK")
# Preprocess images
print("Preprocessing images... ", end="")
preprocessing_pipeline: FunctionList = [
(Preprocessor.bilateral, {
"times": 2
}),
(Preprocessor.sobel, {
"scale": 1,
"delta": 0
}),
]
preprocessed_images = Preprocessor.apply(preprocessing_pipeline, images)
print("OK")
# Calculate asms for initialization
print("Finding initial ASMs... ", end="")
active_shape_models, expected_image_shapes = Incisors.active_shape_models(
preprocessed_images)
print("OK")
# Initialization
print("Initializing ", end="")
initializator: Initializator
if auto:
print("(auto-initialization)... ", end="")
initializator = AutoInitializator()
else:
print("(manual initialization)... ", end="")
initializator = ManualInitializator()
initial_image_shapes = initializator.initialize(active_shape_models,
preprocessed_images)
print("OK")
# Create image shapes with normal images for saving
_, printable_image_shapes = Incisors.active_shape_models(images)
# Evaluation
print("Evaluating...")
evaluator = Evaluator(initial_image_shapes, expected_image_shapes,
printable_image_shapes)
mean_squared_errors = evaluator.quantitative_eval()
print(mean_squared_errors)
evaluator.qualitative_eval()
return
def load_incisor(
incisor=Incisors.UPPER_OUTER_LEFT, extra_text="", blur=False, sobel=False
) -> Tuple[ActiveShapeModel, List[ImageShape]]:
"""Loads asm and imgshapes for sample incisor"""
images = load_images(range(1, 15))
pipeline = [(Preprocessor.bilateral, {"times": 2})]
blurred_images = Preprocessor.apply(pipeline, images)
if blur:
images = blurred_images
sobel_images = apply_sobel(images)
if sobel:
images = sobel_images
asms, imgshapes = Incisors.active_shape_models(images, [incisor])
asm: ActiveShapeModel = asms[incisor]
imgshape: List[ImageShape] = imgshapes[incisor]
return asm, imgshape
def preprocess():
# Load images
print("Loading images... ", end="")
images = load_images(range(1, 15))
print("OK")
# Preprocess images
print("Preprocessing images... ", end="")
preprocessing_pipeline: FunctionList = [
(Preprocessor.bilateral, {
"times": 2
}),
(Preprocessor.sobel, {
"scale": 1,
"delta": 0
}),
]
preprocessed_images = Preprocessor.apply(preprocessing_pipeline, images)
# Plot subplots
f, (ax1, ax2) = plt.subplots(2, 1)
ax1.imshow(images[3], cmap="gray")
ax2.imshow(preprocessed_images[3], cmap="gray")
plt.show()
def test_compile(self):
preprocessor = Preprocessor(COLS_NUM, ["merchantZip"], PATH_READ_X,
PATH_READ_y, DIR_EDA)
def test_normalize(self):
preprocessor = Preprocessor(COLS_NUM, ["merchantZip"], PATH_READ_X,
PATH_READ_y, DIR_EDA)
preprocessor.run(normalize=True)
def test_one_hot_encoding(self):
preprocessor = Preprocessor(COLS_NUM, ["merchantZip"], PATH_READ_X,
PATH_READ_y, DIR_EDA)
preprocessor.run(hot_encode=True)
def test_run(self):
preprocessor = Preprocessor(COLS_NUM, ["merchantZip"], PATH_READ_X,
PATH_READ_y, DIR_EDA)
preprocessor.run(hot_encode=False, normalize=False)
"model_params": {
"n_estimators": [100, 250],
"max_depth": [5, 10, 15]
},
"preprocessing_params": {
"label_encode": True,
"hot_encode": False,
"normalize": True,
"save": False
}
},
}
if __name__ == "__main__":
eda_runner = EDA(COLS_NUM, PATH_READ_X, PATH_READ_y, DIR_EDA)
preprocessor = Preprocessor(COLS_NUM, ["merchantZip"], PATH_READ_X,
PATH_READ_y, DIR_EDA)
eda_runner.run()
# Log results
results = {
"accuracy": [],
"f1": [],
"precision": [],
"recall": [],
"model": []
}
# Iterate over different models
for model_name, model_dict in pipeline_dict.items():
data = preprocessor.run(**model_dict["preprocessing_params"])
# Define X and y