def main():
is_running = True
while is_running:
trainer()
again = get_binary_validation("Do you want to train another model ?",
False)
if not again:
is_running = False
def tester():
existing_files = get_file_list("model", "mdl")
if not existing_files:
print("No models files found. Come back when you have models.")
exit()
target_filename = get_existing_filename(existing_files)
if len(target_filename) == 0:
print("No files selected, quitting...")
exit()
model = None
print(target_filename)
with open("model/" + target_filename, 'rb') as file:
model = pickle.load(file)
is_running = True
while is_running:
raw_sample = get_single_sample()
raw_sample.impostor = True
parser = SampleParser(raw_sample)
timings = parser.timings
timings = timings[:timings[-1]]
try:
results = model.pipeline.predict([timings])
if results[0] == -1:
print('''\u001b[38;5;196m
█████▒▄▄▄ ██▓ ██▓ ▓█████ ▓█████▄
▓██ ▒▒████▄ ▓██▒▓██▒ ▓█ ▀ ▒██▀ ██▌
▒████ ░▒██ ▀█▄ ▒██▒▒██░ ▒███ ░██ █▌
░▓█▒ ░░██▄▄▄▄██ ░██░▒██░ ▒▓█ ▄ ░▓█▄ ▌
░▒█░ ▓█ ▓██▒░██░░██████▒░▒████▒░▒████▓
▒ ░ ▒▒ ▓▒█░░▓ ░ ▒░▓ ░░░ ▒░ ░ ▒▒▓ ▒
░ ▒ ▒▒ ░ ▒ ░░ ░ ▒ ░ ░ ░ ░ ░ ▒ ▒
░ ░ ░ ▒ ▒ ░ ░ ░ ░ ░ ░ ░
░ ░ ░ ░ ░ ░ ░ ░
░
\u001b[0m''')
else:
print('''\u001b[38;5;76m
███████╗██╗ ██╗ ██████╗ ██████╗███████╗███████╗███████╗
██╔════╝██║ ██║██╔════╝██╔════╝██╔════╝██╔════╝██╔════╝
███████╗██║ ██║██║ ██║ █████╗ ███████╗███████╗
╚════██║██║ ██║██║ ██║ ██╔══╝ ╚════██║╚════██║
███████║╚██████╔╝╚██████╗╚██████╗███████╗███████║███████║
╚══════╝ ╚═════╝ ╚═════╝ ╚═════╝╚══════╝╚══════╝╚══════╝
\u001b[0m''')
again = get_binary_validation("Do you want to try again ?", False)
if not again:
is_running = False
except ValueError:
print("It seems that you made a mistake, try again")
continue
def trainer():
existing_files = get_file_list()
if not existing_files:
print("No samples files found. Come back when you have samples.")
exit()
target_filenames = get_existing_filename(existing_files, True)
if len(target_filenames) == 0:
print("No files selected, quitting...")
exit()
sequence = []
for target_filename in target_filenames:
samples = get_sequence_from_file(target_filename)
sequence.extend(samples)
timings_sequences = []
compared_size = None
print("")
for raw_sample in sequence:
parser = SampleParser(raw_sample)
timings_sequences.append(parser.timings)
if compared_size is None:
compared_size = parser.timings[-1]
else:
if parser.timings[-1] != compared_size:
print(
"Error, one sample has a different size ({}), removing it".
format(parser.timings[-1]))
del timings_sequences[-1]
model = Model()
print("{} samples".format(len(timings_sequences)))
# Build the data
trueData = [smp[:smp[-1]] for smp in timings_sequences if smp[-2] == 1]
fakeData = [smp[:smp[-1]] for smp in timings_sequences if smp[-2] == 0]
# Split for training/optimization and final evaluation
train, test = train_test_split(trueData, train_size=0.8, test_size=None)
print("{} samples from user".format(len(trueData)))
print(" {:3d} samples for training".format(len(train)))
print(" {:3d} samples for testing".format(len(test)))
print("{} samples from impostor\n".format(len(fakeData)))
spinner = PixelSpinner("Fitting data to the model... ", )
spinner.start()
# Create a thread for the spinner
t = Thread(target=spinner_loop, args=(spinner, ))
t.do_run = True
t.start()
# Train and optimize
params = Model.findParameters(model, train)
t.do_run = False
t.join()
print("")
# Print a report on the training/optimization phase
# evaluate = Model.evaluate(params["model"], train, test)
# Print a final evaluation of the model agains impostors data
report = Model.report(params["model"], train, test, fakeData)
print_report(report)
save_model = get_binary_validation("Do you want to keep this model ?",
True)
if save_model:
filename = get_custom_filename(target_filenames)
os.makedirs("model", exist_ok=True)
with open("model/" + filename, 'wb') as file:
pickle.dump(params["model"], file, pickle.HIGHEST_PROTOCOL)
print("Model saved in model/" + filename)