def create_mp4(video, audio, save_path=None):
audio = AudioFileClip(audio) if audio is not None else None
fps = video.shape[0] / audio.duration
imageio.mimsave("backend_files/mute.mp4", video)
video = VideoFileClip("backend_files/mute.mp4", audio=False)
video.set_audio(audio)
if save_path is not None:
video.write_videofile(save_path, fps=fps) # , codec='mpeg4')
return video
def form():
## Dealing with form data from the last round, if applicable
try:
classname=request.form['classname']
if request.form['button'] == 'Apply Label':
write_classname = classname
else:
write_classname = request.form['button']
audio_file_id=request.form['audio_file_id']
start_time=request.form['start_time']
output_csv_path='/home/audio_labeler/output_table.csv'
with open(output_csv_path,'a') as fo:
label_duration = 1
fo.write(audio_file_id + ',' + \
str(float(start_time)) + ',' + \
str(float(label_duration)) + ',' + \
'''"''' + write_classname + '''"\n''')
shutil.copy(output_csv_path,'/home/audio_labeler/static/output/output_table.csv')
except:
classname=''
try:
os.remove('/home/audio_labeler/static/' + request.form['temp_wav_filename'])
except:
pass
default_buttons=["Background Speaker","Music","Silence","Multiple Speakers","Not Sure", "Noise",""]
try:
if write_classname not in default_buttons:
label_count_dict[write_classname] += 1
except:
try:
if write_classname not in default_buttons:
label_count_dict[write_classname] = 1
except: pass
label_counts = map(list, label_count_dict.items())
label_counts = sorted(label_counts, key=itemgetter(1))[::-1]
## Launching new round
#audio_filename=random.choice([item for item in os.listdir('/home/audio_labeler/clips') if item[-4:].lower() in ('.mp3','.wav','.mp4')])
media_path = random.choice(media_paths)
audio_file_id = media_path.split('/')[-1][:-4]
duration = media_duration(media_path)
start_time = int((random.random()*(duration-5.0)))
snd = AudioFileClip.AudioFileClip(media_path)
temp_wav_filename = str(random.random())[2:] + '.wav'
snd.subclip(start_time,start_time + 5).write_audiofile('/home/audio_labeler/static/' + temp_wav_filename)
response = render_template('form_audio.html', audio_file_id=audio_file_id, \
start_time=start_time, classname=classname, temp_wav_filename=temp_wav_filename, \
media_path=media_path, label_counts=label_counts[:15])
return response
def subclip(media_path, start_time, end_time, out_dir=''):
if out_dir == '':
out_dir = os.path.dirname(media_path)
audio_filename = media_path.split('/')[-1]
basename = audio_filename[:-4]
start_time = round(float(start_time), 3)
end_time = round(float(end_time), 3)
snd = AudioFileClip.AudioFileClip(media_path)
out_filename = basename + '__' + str(start_time) + '_' + str(
end_time) + '.wav'
snd.subclip(start_time,
end_time).write_audiofile(os.path.join(out_dir, out_filename))
return os.path.join(out_dir, out_filename)
def subclip_list(media_path, range_pairs, out_dir=''):
if out_dir == '':
out_dir = os.path.dirname(media_path)
snd = AudioFileClip.AudioFileClip(media_path)
file_duration = duration(media_path)
for pair in range_pairs:
start = pair[0]
duration_secs = pair[1]
if (float(start) >= 0.0) & (float(duration_secs) > 0.0):
if start + duration_secs > file_duration:
duration_secs = file_duration - start
basename = media_path.split('/')[-1][:-4]
out_filename = basename + '__' + str(round(start, 4)) + '_' + str(
round(duration_secs, 4)) + '.wav'
snd.subclip(float(start),
float(start) + float(duration_secs)).write_audiofile(
os.path.join(out_dir, out_filename))
def read_flute(self):
sigs = []
i = 1
if (self.flute):
print('start flute feature extraction')
os.chdir(os.getcwd() + '\\flute')
for path in os.listdir():
if (i % 100 == 0):
print("flute : ", i, "/", np.size(os.listdir()))
# y, sr = lr.load(path)
signal = sig()
audio = mpe.AudioFileClip(path)
signal.y = np.mean(audio.to_soundarray(buffersize=200), 1)
signal.sr = audio.fps # sampling rate
feat = feature_extraction.features(signal.y, signal.sr)
feat_vec = feat.get_feature_vectors()
self.X.append(feat_vec.tolist())
self.Y.append(2) #2 for flute
sigs.append(signal)
i = i + 1
os.chdir("..")
return sigs
def main(argv):
try:
model_path = sys.argv[1]
resolution_secs = float(sys.argv[2])
dir_path = sys.argv[3]
except getopt.GetoptError:
print "sida_batch_classify.py <pkl model pathname> <directory pathname>"
sys.exit(2)
print('>> Model: ' + model_path)
print('>> Directory: ' + dir_path)
classifier = joblib.load(model_path)
model_basename = os.path.basename(model_path).replace('.pkl','')
media_pathnames = attk.find_media_paths(dir_path)
random.shuffle(media_pathnames)
for media_pathname in media_pathnames:
media_basename = '.'.join(os.path.basename(media_pathname).split('.')[:-1])
csv_path = os.path.join(dir_path, media_basename + '__' + model_basename+'_'+str(resolution_secs)+'.csv')
if not os.path.isfile(csv_path):
tic = timeit.default_timer()
print(' >> Classifying '+ media_pathname.split('/')[-1])
if media_pathname[-4:].lower() == '.wav':
wav_pathname = media_pathname
using_temp_wav = False
else:
wav_pathname = attk.temp_wav_path(media_pathname)
using_temp_wav = True
snd = AudioFileClip.AudioFileClip(wav_pathname)
classifier_output = []
for i in range(int(attk.duration(wav_pathname)/resolution_secs)):
try:
snd.subclip(i * resolution_secs , (i * resolution_secs) + resolution_secs).write_audiofile('/tmp/temp_sida_clip.wav')
classifier_output.append(classify_clip(classifier, '/tmp/temp_sida_clip.wav'))
except Exception as e:
classifier_output.append(0.0)
print(e)
with open(csv_path, 'w') as fo:
i = 0
for value in classifier_output:
start = round(i * resolution_secs, 4)
duration = round(resolution_secs, 4)
fo.write(str(start) + ',' + str(duration) + ',' + str(value) + '\n')
i += 1
if using_temp_wav == True:
os.remove(wav_pathname)
print(" >> Completed in " + str(round(timeit.default_timer() - tic, 4))) + " seconds\n"