def get_symbols(img, model):
masks, coords = get_characters(img)
masks = np.array([dataset_util.reshape(mask, config.SYMBOLS_INPUT_DIM[1:]) for mask in masks])
prediction = classifier.predict(model, masks)
best_pred = classifier_util.get_best_prediction(prediction)
log(f"Symbols: {[classifier.LABELS[x] for x in best_pred]}", LOG_DEBUG, "Symbols")
return best_pred, coords
def get_labels(img, model):
masks, coords = get_characters(img)
masks = array([
dataset_util.reshape(mask, config.CHAR_INPUT_DIM[1:]) for mask in masks
])
prediction = classifier.predict(model, masks)
return [x for x in classifier_util.get_best_prediction(prediction)], coords
def get_indicator_features(image, model):
masks, lit = get_characters(image)
masks = [
dataset_util.reshape(mask, config.CHAR_INPUT_DIM[1:]) for mask in masks
]
prediction = classifier.predict(model, np.array(masks))
best_pred = classifier_util.get_best_prediction(prediction)
return lit, format_label([classifier.LABELS[p] for p in best_pred])
def get_password(img, model):
masks = get_characters(img)
masks = array([
dataset_util.reshape(mask, config.CHAR_INPUT_DIM[1:]) for mask in masks
])
prediction = classifier.predict(model, masks)
best_pred = classifier_util.get_best_prediction(prediction)
characters = [classifier.LABELS[x] for x in best_pred]
return "".join(characters)
def identify_side_features(sides, model):
flattened = []
for side in sides:
for img in side:
converted = dataset_util.reshape(convert_to_cv2(img), config.MODULE_INPUT_DIM[1:])
flattened.append(converted)
preds = module_classifier.predict(model, array(flattened))
predictions = classifier_util.get_best_prediction(preds)
return predictions
def get_serial_number(img, model):
masks, alignment = get_characters(img)
if not alignment:
log("ERROR: Could not determine alignment of serial number",
LOG_WARNING, "Serial Number")
return None
masks = np.array([
dataset_util.reshape(mask, config.CHAR_INPUT_DIM[1:]) for mask in masks
])
prediction = classifier.predict(model, masks)
best_pred = classifier_util.get_best_prediction(prediction)
return create_serial_string(best_pred, alignment) # Return actual string.
def get_details_async(model, holder):
duration_masks = get_duration_characters()
if len(duration_masks) != 3:
log(f"WARNING: Bomb duration string length != 3 (len={len(get_duration_characters)}).",
config.LOG_WARNING)
module_masks = get_module_characters()
masks = duration_masks + module_masks
masks = np.array([dataset_util.reshape(mask, config.CHAR_INPUT_DIM[1:]) for mask in masks])
prediction = classifier.predict(model, masks)
best_pred = classifier_util.get_best_prediction(prediction)
labels = [classifier.LABELS[p] for p in best_pred]
holder.append(format_time(labels[:3]))
holder.append(fix_number(labels[3:5]))
def get_button_features(image, model):
masks, color = get_characters(image)
if len(masks) == 4:
return "Hold", color
elif len(masks) == 8:
return "Detonate", color
else:
masks = np.array(
[dataset_util.reshape(masks[0], config.CHAR_INPUT_DIM[1:])])
prediction = classifier.predict(model, masks[0])
best_pred = classifier_util.get_best_prediction(prediction)
if classifier.LABELS[best_pred[0]] == "a":
return "Abort", color
return "Press", color
def get_words(img, model):
_, words, coords = get_characters(img)
predictions = []
for masks in words:
result = " "
if masks is not None:
masks = np.array([
dataset_util.reshape(mask, config.CHAR_INPUT_DIM[1:])
for mask in masks
])
prediction = classifier.predict(model, masks)
result = get_word(prediction)
predictions.append(result)
return predictions, coords
def process_bomb_data(images):
IMAGES_CAPTURED = 0
predictions = []
for img in images:
cv2_img = convert_to_cv2(img)
path = f"../resources/training_images/"
if AUTO_LABEL:
resized = dataset_util.reshape(cv2_img, config.MODULE_INPUT_DIM[1:])
pred = classifier.predict(MODEL, resized)
label = classifier_util.get_best_prediction(pred)[0]
predictions.append(label)
if label in INCLUDED_LABELS:
if label < 7:
label_name = clean_file_path(classifier.LABELS[label])
path = f"{path}{label_name}/"
if DATA_TYPE != "modules":
num_images = len(glob(path + "*.png"))
imwrite(f"{path}{num_images:03d}.png", cv2_img)
IMAGES_CAPTURED += 1
log(f"Captured {IMAGES_CAPTURED} bomb images.")
return predictions