def main()

in evaluate.py [0:0]

• 31 lines of code
• 3 McCabe index (conditional complexity)

def main():
	parser = argparse.ArgumentParser()
	parser.add_argument('--results_root', type=str, required=True)
	parser.add_argument('--audio_sampling_rate', default=16000, type=int, help='audio sampling rate')
	parser.add_argument('--real_mono', default=False, type=bool, help='whether the input predicted binaural audio is mono audio')
	parser.add_argument('--normalization', default=False, type=bool)
	args = parser.parse_args()
	stft_distance_list = []
	envelope_distance_list = []

	audioNames = os.listdir(args.results_root)
	index = 1
	for audio_name in audioNames:
		if index % 10 == 0:
			print "Evaluating testing example " + str(index) + " :", audio_name
		#check whether input binaural is mono, replicate to two channels if it's mono
		if args.real_mono:
			mono_sound, audio_rate = librosa.load(os.path.join(args.results_root, audio_name, 'mixed_mono.wav'), sr=args.audio_sampling_rate)
			predicted_binaural = np.repeat(np.expand_dims(mono_sound, 0), 2, axis=0)
			if args.normalization:
				predicted_binaural = normalize(predicted_binaural)
		else:
			predicted_binaural, audio_rate = librosa.load(os.path.join(args.results_root, audio_name, 'predicted_binaural.wav'), sr=args.audio_sampling_rate, mono=False)
			if args.normalization:
				predicted_binaural = normalize(predicted_binaural)
		gt_binaural, audio_rate = librosa.load(os.path.join(args.results_root, audio_name, 'input_binaural.wav'), sr=args.audio_sampling_rate, mono=False)
		if args.normalization:
			gt_binaural = normalize(gt_binaural)
		
		#get results for this audio
		stft_distance_list.append(STFT_L2_distance(predicted_binaural, gt_binaural))
		envelope_distance_list.append(Envelope_distance(predicted_binaural, gt_binaural))
		index = index + 1

	#print the results
	print "STFT L2 Distance: ", stat.mean(stft_distance_list), stat.stdev(stft_distance_list), stat.stdev(stft_distance_list) / np.sqrt(len(stft_distance_list))
	print "Average Envelope Distance: ", stat.mean(envelope_distance_list), stat.stdev(envelope_distance_list), stat.stdev(envelope_distance_list) / np.sqrt(len(envelope_distance_list))