def apply_rast_tshd(image, seed, tshd_val):
seed_image = x_test[int(seed)]
xml_desc = vectorization_tools.vectorize(seed_image)
seed = rasterization_tools.rasterize_in_memory(xml_desc)
distance_seed = get_distance(seed, image)
print("RAST %s" % distance_seed)
if distance_seed < tshd_val:
return True
else:
return False
def mutate(self):
condition = True
counter_mutations = 0
while condition:
# Select mutation operator.
rand_mutation_probability = random.uniform(0, 1)
rand_mutation_prob = random.uniform(0, 1)
if rand_mutation_probability >= MUTOPPROB:
if rand_mutation_prob >= MUTOFPROB:
mutation = 1
else:
mutation = 2
else:
if rand_mutation_prob >= MUTOFPROB:
mutation = 3
else:
mutation = 4
counter_mutations += 1
mutant_vector = mutation_manager.mutate(self.digit.xml_desc, mutation, counter_mutations/20)
mutant_xml_desc = vectorization_tools.create_svg_xml(mutant_vector)
rasterized_digit = rasterization_tools.rasterize_in_memory(mutant_xml_desc)
distance_inputs = get_distance(self.digit.purified, rasterized_digit)
if (TSHD_TYPE == '0'):
if distance_inputs != 0:
condition = False
elif (TSHD_TYPE == '1'):
seed_image = DigitMutator.x_test[int(self.digit.seed)]
xml_desc = vectorization_tools.vectorize(seed_image)
seed = rasterization_tools.rasterize_in_memory(xml_desc)
distance_seed = get_distance(seed, rasterized_digit)
if distance_inputs != 0 and distance_seed <= DISTANCE and distance_seed != 0:
condition = False
elif (TSHD_TYPE == '2'):
seed = reshape(DigitMutator.x_test[int(self.digit.seed)])
distance_seed = get_distance(seed, rasterized_digit)
if distance_inputs != 0 and distance_seed <= DISTANCE_SEED and distance_seed != 0:
condition = False
self.digit.xml_desc = mutant_xml_desc
self.digit.purified = rasterized_digit
self.digit.predicted_label = None
self.digit.confidence = None
def __init__(self, basepath):
super(DeepHyperionSample, self).__init__(basepath)
# Load the metadata (json) file and the files to compute the features (xml, svg)
npy_path = basepath + ".npy"
json_path = basepath + ".json"
svg_path = basepath + ".svg"
with open(json_path) as jf:
json_data = json.load(jf)
self.image = np.load(npy_path)
if os.path.exists(svg_path):
self.approximated_model = False
with open(svg_path, 'r') as input_file:
self.xml_desc = input_file.read()
else:
self.approximated_model = True
self.xml_desc = vectorize(self.image)
# Set Sample attributes
self.id = json_data["id"]
self.tool = json_data["tool"]
self.misbehaviour = json_data["misbehaviour"]
self.run = json_data["run"]
self.timestamp = json_data["timestamp"]
self.elapsed = json_data["elapsed"]
# Compute the metrics, those are not available in the metadata json
self.features["moves"] = metrics.move_distance(self.xml_desc)
self.features["bitmaps"] = metrics.dark_bitmaps(self.image)
self.features["orientation"] = metrics.orientation_calc(self.image)
# Set MNistSample attributes
self.performance = json_data["performance"]
self.predicted_label = json_data["predicted_label"]
self.expected_label = json_data["expected_label"]
self.seed = json_data["seed"]
# Store to file besides the input json
self.dump()
def __init__(self, basepath):
super(DLFuzzSample, self).__init__(basepath)
# Load the metadata (json) file and the files to compute the features (xml, npy)
npy_path = basepath + ".npy"
# TODO Does this have to be an instance attribute? Maybe we can remove it after we compute the features?
self.image = np.load(npy_path)
json_path = basepath + ".json"
with open(json_path) as jf:
json_data = json.load(jf)
try:
# TODO Does this have to be an instance attribute? Maybe we can remove it after we compute the features?
self.xml_desc = json_data["xml_desc"]
# TODO Does this have to be an instance attribute? Maybe we can remove it after we compute the features?
self.approximated_model = False
except KeyError:
# TODO Does this have to be an instance attribute? Maybe we can remove it after we compute the features?
self.xml_desc = vectorize(self.image)
# TODO Does this have to be an instance attribute? Maybe we can remove it after we compute the features?
self.approximated_model = True
# Set Sample attributes
self.id = json_data["id"]
self.tool = json_data["tool"]
self.misbehaviour = json_data["misbehaviour"]
self.run = json_data["run"]
self.timestamp = json_data["timestamp"]
self.elapsed = json_data["elapsed"]
# Compute the metrics, those are not available in the metadata json
self.features["moves"] = metrics.move_distance(self.xml_desc)
self.features["bitmaps"] = metrics.dark_bitmaps(self.image)
self.features["orientation"] = metrics.orientation_calc(self.image)
# Set MNistSample attributes
self.performance = json_data["performance"]
self.predicted_label = json_data["predicted_label"]
self.expected_label = json_data["expected_label"]
self.seed = json_data["seed"]
# Store to file besides the input json
self.dump()
def generate_digit(seed):
seed_image = x_test[int(seed)]
xml_desc = vectorization_tools.vectorize(seed_image)
return Digit(xml_desc, EXPECTED_LABEL, seed)