opt = Adam(lr=0.1, beta_1=0.9, beta_2=0.999)
model = Sequential()
model.add(Dense(250, input_shape=[5, 199], activation='relu'))
model.add(Dense(120, activation='relu'))
model.add(Flatten())
#model.add(Dropout(rate=0.25))
model.add(Dense(5, activation='softmax'))
model.load_weights('nn_weights_5000_5_parameters.h5')
model.compile(optimizer=opt, loss='categorical_crossentropy')
audio_file = sys.argv[1]
print(audio_file)
file_freq_peaks = check_peaks.check_peaks(sys.argv[1])
file_phases = check_phase.check_phase(sys.argv[1])
file_samples_peaks = check_samples.check_peaks_sample(sys.argv[1])
file_signal_noise_ratio = check_signal_to_noise_ratio.check_signal_to_noise(
sys.argv[1])
file_odd_numbers = check_odd_numbers.check_odd_numbers(sys.argv[1])
input_data = [
file_freq_peaks, file_phases, file_samples_peaks, file_signal_noise_ratio,
file_odd_numbers
]
input_data = np.array([input_data], dtype=np.float32)
print(model.predict(input_data))
file_samples_peaks = []
file_phases = []
file_types = []
file_samples_signal_noise = []
file_odd_numbers = []
dir_counter = 0
counter = 0
for directory_name in directories[0]:
dir_counter += 1
for audio_file in audio_files[dir_counter]:
file_freq_peaks.append(check_peaks.check_peaks(original_path + directory_name + '/' + audio_file))
file_phases.append(check_phase.check_phase(original_path + directory_name + '/' + audio_file))
file_samples_peaks.append(check_samples.check_peaks_sample(original_path + directory_name + '/' + audio_file))
file_samples_signal_noise.append(check_signal_to_noise_ratio.check_signal_to_noise(original_path + directory_name + '/' + audio_file))
file_odd_numbers.append(check_odd_numbers.check_odd_numbers(original_path + directory_name + '/' + audio_file))
if 'audio in' in directory_name:
file_types.append([1, 0, 0, 0, 0])
elif 'Echo Hiding' in directory_name:
file_types.append([0, 1, 0, 0, 0])
elif 'LSB' in directory_name:
file_types.append([0, 0, 1, 0, 0])
elif 'Phase Coding' in directory_name:
file_types.append([0, 0, 0, 1, 0])
elif 'Spread Spectrum' in directory_name:
file_types.append([0, 0, 0, 0, 1])
counter += 1
percentage_done = (counter/(len(directories[0]) * len(audio_files[1])))*100
# input_data = np.array([input_data], dtype=np.float32)
# print(model.predict(input_data))
counter = 0
audio_in = [1, 0, 0, 0, 0]
echo_hiding = [0, 1, 0, 0, 0]
lsb = [0, 0, 1, 0, 0]
phase_coding = [0, 0, 0, 1, 0]
spread_spectrum = [0, 0, 0, 0, 1]
for audio_file in audio_files[0]:
file_freq_peaks = check_peaks.check_peaks(original_path + '/' + audio_file)
file_phases = check_phase.check_phase(original_path + '/' + audio_file)
file_samples_peaks = check_samples.check_peaks_sample(original_path + '/' +
audio_file)
file_signal_to_noise = check_signal_to_noise_ratio.check_signal_to_noise(
original_path + '/' + audio_file)
file_odd_numbers = check_odd_numbers.check_odd_numbers(original_path +
'/' + audio_file)
input_data = [
file_freq_peaks, file_phases, file_samples_peaks, file_signal_to_noise,
file_odd_numbers
]
input_data = np.array([input_data], dtype=np.float32)
# print(input_data.shape)
predicted_results = model.predict(input_data)
real_result = np.array(audio_in)
if real_result.argmax() == predicted_results.argmax():
import check_samples
print(check_samples.check_peaks_sample('wav_stego/audio in/ballad.wav'))
array1 = [1, 0, 0, 0]
array2 = [1, 0, 0, 0]
print(array1 == array2)