def main(args):
# get parsed arguments from user
input_dir = args.input_dir
architecture = args.architecture
color_mode = args.color
loss = args.loss
batch_size = args.batch
epochs = args.epochs
lr_estimate = args.lr_estimate
policy = args.policy
# check arguments
check_arguments(architecture, color_mode, loss)
# get autoencoder
autoencoder = AutoEncoder(input_dir, architecture, color_mode, loss,
batch_size)
# load data as generators that yield batches of preprocessed images
preprocessor = Preprocessor(
input_directory=input_dir,
rescale=autoencoder.rescale,
shape=autoencoder.shape,
color_mode=autoencoder.color_mode,
)
train_generator = preprocessor.get_train_generator(
batch_size=autoencoder.batch_size, shuffle=True)
validation_generator = preprocessor.get_val_generator(
batch_size=autoencoder.batch_size,
shuffle=False,
purpose="val",
)
# find best learning rates for training
lr_opt = autoencoder.find_lr_opt(train_generator, validation_generator,
lr_estimate)
# train with optimal learning rate
autoencoder.fit(lr_opt=lr_opt, epochs=epochs, policy=policy)
# save model and configuration
autoencoder.save()
# inspect validation and test images for visual assessement
if args.inspect:
inspection.inspect_images(model_path=autoencoder.save_path)
logger.info("done.")
return
def main(args):
# Get validation arguments
model_path = args.path
method = args.method
dtype = args.dtype
# ============= LOAD MODEL AND PREPROCESSING CONFIGURATION ================
# load model and info
model, info, _ = utils.load_model_HDF5(model_path)
# set parameters
input_directory = info["data"]["input_directory"]
architecture = info["model"]["architecture"]
loss = info["model"]["loss"]
rescale = info["preprocessing"]["rescale"]
shape = info["preprocessing"]["shape"]
color_mode = info["preprocessing"]["color_mode"]
vmin = info["preprocessing"]["vmin"]
vmax = info["preprocessing"]["vmax"]
nb_validation_images = info["data"]["nb_validation_images"]
# get the correct preprocessing function
preprocessing_function = get_preprocessing_function(architecture)
# ========= LOAD AND PREPROCESS VALIDATION & FINETUNING IMAGES =============
# initialize preprocessor
preprocessor = Preprocessor(
input_directory=input_directory,
rescale=rescale,
shape=shape,
color_mode=color_mode,
preprocessing_function=preprocessing_function,
)
# -------------------------------------------------------------------
# get validation generator
validation_generator = preprocessor.get_val_generator(
batch_size=nb_validation_images, shuffle=False)
# retrieve preprocessed validation images from generator
imgs_val_input = validation_generator.next()[0]
# retrieve validation image_names
filenames_val = validation_generator.filenames
# reconstruct (i.e predict) validation images
imgs_val_pred = model.predict(imgs_val_input)
# convert to grayscale if RGB
if color_mode == "rgb":
imgs_val_input = tf.image.rgb_to_grayscale(imgs_val_input).numpy()
imgs_val_pred = tf.image.rgb_to_grayscale(imgs_val_pred).numpy()
# remove last channel since images are grayscale
imgs_val_input = imgs_val_input[:, :, :, 0]
imgs_val_pred = imgs_val_pred[:, :, :, 0]
# instantiate TensorImages object to compute validation resmaps
tensor_val = resmaps.TensorImages(
imgs_input=imgs_val_input,
imgs_pred=imgs_val_pred,
vmin=vmin,
vmax=vmax,
method=method,
dtype=dtype,
filenames=filenames_val,
)
# -------------------------------------------------------------------
# get finetuning generator
nb_test_images = preprocessor.get_total_number_test_images()
finetuning_generator = preprocessor.get_finetuning_generator(
batch_size=nb_test_images, shuffle=False)
# retrieve preprocessed test images from generator
imgs_test_input = finetuning_generator.next()[0]
filenames_test = finetuning_generator.filenames
# select a representative subset of test images for finetuning
# using stratified sampling
assert "good" in finetuning_generator.class_indices
index_array = finetuning_generator.index_array
classes = finetuning_generator.classes
_, index_array_ft, _, classes_ft = train_test_split(
index_array,
classes,
test_size=FINETUNE_SPLIT,
random_state=42,
stratify=classes,
)
# get correct classes corresponding to selected images
good_class_i = finetuning_generator.class_indices["good"]
y_ft_true = np.array(
[0 if class_i == good_class_i else 1 for class_i in classes_ft])
# select test images for finetuninig
imgs_ft_input = imgs_test_input[index_array_ft]
filenames_ft = list(np.array(filenames_test)[index_array_ft])
# reconstruct (i.e predict) finetuning images
imgs_ft_pred = model.predict(imgs_ft_input)
# convert to grayscale if RGB
if color_mode == "rgb":
imgs_ft_input = tf.image.rgb_to_grayscale(imgs_ft_input).numpy()
imgs_ft_pred = tf.image.rgb_to_grayscale(imgs_ft_pred).numpy()
# remove last channel since images are grayscale
imgs_ft_input = imgs_ft_input[:, :, :, 0]
imgs_ft_pred = imgs_ft_pred[:, :, :, 0]
# instantiate TensorImages object to compute finetuning resmaps
tensor_ft = resmaps.TensorImages(
imgs_input=imgs_ft_input,
imgs_pred=imgs_ft_pred,
vmin=vmin,
vmax=vmax,
method=method,
dtype=dtype,
filenames=filenames_ft,
)
# ======================== COMPUTE THRESHOLDS ===========================
# initialize finetuning dictionary
dict_finetune = {
"min_area": [],
"threshold": [],
"TPR": [],
"TNR": [],
"FPR": [],
"FNR": [],
"score": [],
}
# create discrete min_area values
min_areas = np.arange(start=5, stop=505, step=STEP_MIN_AREA)
length = len(min_areas)
for i, min_area in enumerate(min_areas):
print("step {}/{} | current min_area = {}".format(
i + 1, length, min_area))
# compute threshold corresponding to current min_area
threshold = determine_threshold(
resmaps=tensor_val.resmaps,
min_area=min_area,
thresh_min=tensor_val.thresh_min,
thresh_max=tensor_val.thresh_max,
thresh_step=tensor_val.thresh_step,
)
# apply the min_area, threshold pair to finetuning images
y_ft_pred = predict_classes(resmaps=tensor_ft.resmaps,
min_area=min_area,
threshold=threshold)
# confusion matrix
tnr, fpr, fnr, tpr = confusion_matrix(y_ft_true,
y_ft_pred,
normalize="true").ravel()
# record current results
dict_finetune["min_area"].append(min_area)
dict_finetune["threshold"].append(threshold)
dict_finetune["TPR"].append(tpr)
dict_finetune["TNR"].append(tnr)
dict_finetune["FPR"].append(fpr)
dict_finetune["FNR"].append(fnr)
dict_finetune["score"].append((tpr + tnr) / 2)
# get min_area, threshold pair corresponding to best score
max_score_i = np.argmax(dict_finetune["score"])
max_score = float(dict_finetune["score"][max_score_i])
best_min_area = int(dict_finetune["min_area"][max_score_i])
best_threshold = float(dict_finetune["threshold"][max_score_i])
# ===================== SAVE VALIDATION RESULTS ========================
# create a results directory if not existent
model_dir_name = os.path.basename(str(Path(model_path).parent))
save_dir = os.path.join(
os.getcwd(),
"results",
input_directory,
architecture,
loss,
model_dir_name,
"finetuning",
"{}_{}".format(method, dtype),
)
if not os.path.isdir(save_dir):
os.makedirs(save_dir)
# save area and threshold pair
finetuning_result = {
"best_min_area": best_min_area,
"best_threshold": best_threshold,
"best_score": max_score,
"method": method,
"dtype": dtype,
"split": FINETUNE_SPLIT,
}
print("finetuning results: {}".format(finetuning_result))
# save validation result
with open(os.path.join(save_dir, "finetuning_result.json"),
"w") as json_file:
json.dump(finetuning_result, json_file, indent=4, sort_keys=False)
logger.info("finetuning results saved at {}".format(save_dir))
# save finetuning plots
plot_min_area_threshold(dict_finetune,
index_best=max_score_i,
save_dir=save_dir)
plot_scores(dict_finetune, index_best=max_score_i, save_dir=save_dir)
return
def main(args):
# Get validation arguments
model_path = args.path
color = args.color # NOT YET TAKEN INTO ACCOUNT
method = args.method
dtype = args.dtype
# ============= LOAD MODEL AND PREPROCESSING CONFIGURATION ================
# load model and info
model, info, _ = utils.load_model_HDF5(model_path)
# set parameters
input_directory = info["data"]["input_directory"]
architecture = info["model"]["architecture"]
loss = info["model"]["loss"]
rescale = info["preprocessing"]["rescale"]
shape = info["preprocessing"]["shape"]
color_mode = info["preprocessing"]["color_mode"]
vmin = info["preprocessing"]["vmin"]
vmax = info["preprocessing"]["vmax"]
nb_validation_images = info["data"]["nb_validation_images"]
# ========= LOAD AND PREPROCESS VALIDATION & FINETUNING IMAGES =============
# initialize preprocessor
preprocessor = Preprocessor(
input_directory=input_directory,
rescale=rescale,
shape=shape,
color_mode=color_mode,
)
# -------------------------------------------------------------------
# get validation generator
validation_generator = preprocessor.get_val_generator(
batch_size=nb_validation_images, shuffle=False)
# retrieve preprocessed validation images from generator
imgs_val_input = validation_generator.next()[0]
# retrieve validation image_names
filenames_val = validation_generator.filenames
# reconstruct (i.e predict) validation images
imgs_val_pred = model.predict(imgs_val_input)
# instantiate TensorImages object to compute validation resmaps
tensor_val = postprocessing.TensorImages(
imgs_input=imgs_val_input,
imgs_pred=imgs_val_pred,
vmin=vmin,
vmax=vmax,
color="grayscale",
method=method,
dtype=dtype,
filenames=filenames_val,
)
# -------------------------------------------------------------------
# get finetuning generator
nb_test_images = preprocessor.get_total_number_test_images()
finetuning_generator = preprocessor.get_finetuning_generator(
batch_size=nb_test_images, shuffle=False)
# retrieve preprocessed test images from generator
imgs_test_input = finetuning_generator.next()[0]
filenames_test = finetuning_generator.filenames
# select a representative subset of test images for finetuning
# using stratified sampling
assert "good" in finetuning_generator.class_indices
index_array = finetuning_generator.index_array
classes = finetuning_generator.classes
_, index_array_ft, _, classes_ft = train_test_split(
index_array,
classes,
test_size=FINETUNE_SPLIT,
random_state=42,
stratify=classes,
)
# get correct classes corresponding to selected images
good_class_i = finetuning_generator.class_indices["good"]
y_ft_true = np.array(
[0 if class_i == good_class_i else 1 for class_i in classes_ft])
# select test images for finetuninig
imgs_ft_input = imgs_test_input[index_array_ft]
filenames_ft = list(np.array(filenames_test)[index_array_ft])
# reconstruct (i.e predict) finetuning images
imgs_ft_pred = model.predict(imgs_ft_input)
# instantiate TensorImages object to compute finetuning resmaps
tensor_ft = postprocessing.TensorImages(
imgs_input=imgs_ft_input,
imgs_pred=imgs_ft_pred,
vmin=vmin,
vmax=vmax,
color="grayscale",
method=method,
dtype=dtype,
filenames=filenames_ft,
)
# ======================== COMPUTE THRESHOLDS ===========================
# initialize finetuning dictionary
dict_finetune = {
"min_area": [],
"threshold": [],
"TPR": [],
"TNR": [],
"FPR": [],
"FNR": [],
"score": [],
}
# initialize discrete min_area values
min_areas = np.arange(
start=START_MIN_AREA,
stop=STOP_MIN_AREA,
step=STEP_MIN_AREA,
)
# initialize thresholds
thresholds = np.arange(
start=tensor_val.thresh_min,
stop=tensor_val.thresh_max + tensor_val.thresh_step,
step=tensor_val.thresh_step,
)
# compute largest anomaly areas in resmaps for increasing thresholds
logger.info(
"step 1/2: computing largest anomaly areas for increasing thresholds..."
)
largest_areas = calculate_largest_areas(
resmaps=tensor_val.resmaps,
thresholds=thresholds,
)
# select best minimum area and threshold pair to use for testing
logger.info(
"step 2/2: selecting best minimum area and threshold pair for testing..."
)
printProgressBar(0,
len(min_areas),
prefix="Progress:",
suffix="Complete",
length=80)
for i, min_area in enumerate(min_areas):
# compare current min_area with the largest area
for index, largest_area in enumerate(largest_areas):
if min_area > largest_area:
break
# select threshold corresponding to current min_area
threshold = thresholds[index]
# apply the min_area, threshold pair to finetuning images
y_ft_pred = predict_classes(resmaps=tensor_ft.resmaps,
min_area=min_area,
threshold=threshold)
# confusion matrix
tnr, fpr, fnr, tpr = confusion_matrix(y_ft_true,
y_ft_pred,
normalize="true").ravel()
# record current results
dict_finetune["min_area"].append(min_area)
dict_finetune["threshold"].append(threshold)
dict_finetune["TPR"].append(tpr)
dict_finetune["TNR"].append(tnr)
dict_finetune["FPR"].append(fpr)
dict_finetune["FNR"].append(fnr)
dict_finetune["score"].append((tpr + tnr) / 2)
# print progress bar
printProgressBar(i + 1,
len(min_areas),
prefix="Progress:",
suffix="Complete",
length=80)
# get min_area, threshold pair corresponding to best score
max_score_i = np.argmax(dict_finetune["score"])
max_score = float(dict_finetune["score"][max_score_i])
best_min_area = int(dict_finetune["min_area"][max_score_i])
best_threshold = float(dict_finetune["threshold"][max_score_i])
# ===================== SAVE FINETUNING RESULTS ========================
# create a results directory if not existent
model_dir_name = os.path.basename(str(Path(model_path).parent))
save_dir = os.path.join(
os.getcwd(),
"results",
input_directory,
architecture,
loss,
model_dir_name,
"finetuning",
"{}_{}".format(method, dtype),
)
if not os.path.isdir(save_dir):
os.makedirs(save_dir)
# save area and threshold pair
finetuning_result = {
"best_min_area": best_min_area,
"best_threshold": best_threshold,
"best_score": max_score,
"method": method,
"dtype": dtype,
"split": FINETUNE_SPLIT,
}
logger.info("finetuning results: {}".format(finetuning_result))
# save validation result
with open(os.path.join(save_dir, "finetuning_result.json"),
"w") as json_file:
json.dump(finetuning_result, json_file, indent=4, sort_keys=False)
logger.info("finetuning results saved at {}".format(save_dir))
# save finetuning plots
plot_min_area_threshold(dict_finetune,
index_best=max_score_i,
save_dir=save_dir)
plot_scores(dict_finetune, index_best=max_score_i, save_dir=save_dir)
return
def main(args):
# get parsed arguments from user
input_dir = args.input_dir
architecture = args.architecture
color_mode = args.color
loss = args.loss
batch_size = args.batch
# check arguments
check_arguments(architecture, color_mode, loss)
# get autoencoder
autoencoder = AutoEncoder(input_dir, architecture, color_mode, loss,
batch_size)
# load data as generators that yield batches of preprocessed images
preprocessor = Preprocessor(
input_directory=input_dir,
rescale=autoencoder.rescale,
shape=autoencoder.shape,
color_mode=autoencoder.color_mode,
preprocessing_function=autoencoder.preprocessing_function,
)
train_generator = preprocessor.get_train_generator(
batch_size=autoencoder.batch_size, shuffle=True)
validation_generator = preprocessor.get_val_generator(
batch_size=autoencoder.batch_size, shuffle=True)
# find best learning rates for training
autoencoder.find_opt_lr(train_generator, validation_generator)
# train
autoencoder.fit()
# save model
autoencoder.save()
if args.inspect:
# -------------- INSPECTING VALIDATION IMAGES --------------
logger.info("generating inspection plots of validation images...")
# create a directory to save inspection plots
inspection_val_dir = os.path.join(autoencoder.save_dir,
"inspection_val")
if not os.path.isdir(inspection_val_dir):
os.makedirs(inspection_val_dir)
inspection_val_generator = preprocessor.get_val_generator(
batch_size=autoencoder.learner.val_data.samples, shuffle=False)
imgs_val_input = inspection_val_generator.next()[0]
filenames_val = inspection_val_generator.filenames
# get reconstructed images (i.e predictions) on validation dataset
logger.info("reconstructing validation images...")
imgs_val_pred = autoencoder.model.predict(imgs_val_input)
# convert to grayscale if RGB
if color_mode == "rgb":
imgs_val_input = tf.image.rgb_to_grayscale(imgs_val_input).numpy()
imgs_val_pred = tf.image.rgb_to_grayscale(imgs_val_pred).numpy()
# remove last channel since images are grayscale
imgs_val_input = imgs_val_input[:, :, :, 0]
imgs_val_pred = imgs_val_pred[:, :, :, 0]
# instantiate TensorImages object to compute validation resmaps
tensor_val = postprocessing.TensorImages(
imgs_input=imgs_val_input,
imgs_pred=imgs_val_pred,
vmin=autoencoder.vmin,
vmax=autoencoder.vmax,
method=autoencoder.loss,
dtype="float64",
filenames=filenames_val,
)
# generate and save inspection validation plots
tensor_val.generate_inspection_plots(group="validation",
save_dir=inspection_val_dir)
# -------------- INSPECTING TEST IMAGES --------------
logger.info("generating inspection plots of test images...")
# create a directory to save inspection plots
inspection_test_dir = os.path.join(autoencoder.save_dir,
"inspection_test")
if not os.path.isdir(inspection_test_dir):
os.makedirs(inspection_test_dir)
nb_test_images = preprocessor.get_total_number_test_images()
inspection_test_generator = preprocessor.get_test_generator(
batch_size=nb_test_images, shuffle=False)
imgs_test_input = inspection_test_generator.next()[0]
filenames_test = inspection_test_generator.filenames
# get reconstructed images (i.e predictions) on validation dataset
logger.info("reconstructing test images...")
imgs_test_pred = autoencoder.model.predict(imgs_test_input)
# convert to grayscale if RGB
if color_mode == "rgb":
imgs_test_input = tf.image.rgb_to_grayscale(
imgs_test_input).numpy()
imgs_test_pred = tf.image.rgb_to_grayscale(imgs_test_pred).numpy()
# remove last channel since images are grayscale
imgs_test_input = imgs_test_input[:, :, :, 0]
imgs_test_pred = imgs_test_pred[:, :, :, 0]
# instantiate TensorImages object to compute test resmaps
tensor_test = postprocessing.TensorImages(
imgs_input=imgs_test_input,
imgs_pred=imgs_test_pred,
vmin=autoencoder.vmin,
vmax=autoencoder.vmax,
method=autoencoder.loss,
dtype="float64",
filenames=filenames_test,
)
# generate and save inspection test plots
tensor_test.generate_inspection_plots(group="test",
save_dir=inspection_test_dir)
logger.info("done.")
return
def inspect_images(model_path):
# load model for inspection
logger.info("loading model for inspection...")
model, info, _ = utils.load_model_HDF5(model_path)
save_dir = os.path.dirname(model_path)
input_dir = info["data"]["input_directory"]
# architecture = info["model"]["architecture"]
# loss = info["model"]["loss"]
rescale = info["preprocessing"]["rescale"]
shape = info["preprocessing"]["shape"]
color_mode = info["preprocessing"]["color_mode"]
vmin = info["preprocessing"]["vmin"]
vmax = info["preprocessing"]["vmax"]
nb_validation_images = info["data"]["nb_validation_images"]
# instantiate preprocessor object to preprocess validation and test inspection images
preprocessor = Preprocessor(
input_directory=input_dir,
rescale=rescale,
shape=shape,
color_mode=color_mode,
)
# -------------- INSPECTING VALIDATION IMAGES --------------
logger.info("generating inspection plots for validation images...")
inspection_val_generator = preprocessor.get_val_generator(
batch_size=nb_validation_images, shuffle=False)
imgs_val_input = inspection_val_generator.next()[0]
filenames_val = inspection_val_generator.filenames
# get indices of validation inspection images
val_insp_i = [
filenames_val.index(filename)
for filename in config.FILENAMES_VAL_INSPECTION
]
imgs_val_input = imgs_val_input[val_insp_i]
# reconstruct validation inspection images (i.e predict)
imgs_val_pred = model.predict(imgs_val_input)
# instantiate ResmapPlotter object to compute resmaps
postproc_val = postprocessing.ResmapPlotter(
imgs_input=imgs_val_input,
imgs_pred=imgs_val_pred,
filenames=config.FILENAMES_VAL_INSPECTION,
color="grayscale",
vmin=vmin,
vmax=vmax,
)
# generate resmaps and save
fig_res_val = postproc_val.generate_inspection_figure()
fig_res_val.savefig(os.path.join(save_dir, "fig_insp_val.svg"))
# -------------- INSPECTING TEST IMAGES --------------
logger.info("generating inspection plots for test images...")
nb_test_images = preprocessor.get_total_number_test_images()
inspection_test_generator = preprocessor.get_test_generator(
batch_size=nb_test_images, shuffle=False)
# get preprocessed test images
imgs_test_input = inspection_test_generator.next()[0]
filenames_test = inspection_test_generator.filenames
# get indices of test inspection images
test_insp_i = [
filenames_test.index(filename)
for filename in config.FILENAMES_TEST_INSPECTION
]
imgs_test_input = imgs_test_input[test_insp_i]
# reconstruct inspection test images (i.e predict)
imgs_test_pred = model.predict(imgs_test_input)
# instantiate ResmapPlotter object to compute resmaps
postproc_test = postprocessing.ResmapPlotter(
imgs_input=imgs_test_input,
imgs_pred=imgs_test_pred,
filenames=config.FILENAMES_TEST_INSPECTION,
color="grayscale",
vmin=vmin,
vmax=vmax,
)
# generate resmaps and save
fig_res_test = postproc_test.generate_inspection_figure()
fig_res_test.savefig(os.path.join(save_dir, "fig_insp_test.svg"))
# --------------------------------------------------
# fig_score_insp = postproc_test.generate_score_scatter_plot(
# inspection_test_generator, model_path, filenames_test_insp
# )
# fig_score_insp.savefig(os.path.join(save_dir, "fig_score_insp.svg"))
# fig_score_test = postproc_test.generate_score_scatter_plot(
# inspection_test_generator, model_path
# )
# fig_score_test.savefig(os.path.join(save_dir, "fig_score_test.svg"))
return