def generate_data(sub_names, mov_names):
speech = []
noise = []
for sub_name, mov_name in zip(sub_names, mov_names):
fsub = open(sub_name)
intervels = []
while True:
if fsub.readline().strip() == '[Events]':
break
for line in fsub:
if line.startswith('Dialogue:'):
line = line.strip().split(',')
if line[3] != 'Default' or line[9].startswith('{'):
intervels.append(
(parse_time(line[1]), parse_time(line[2]), False))
else:
intervels.append(
(parse_time(line[1]), parse_time(line[2]), True))
intervels.sort(cmp=lambda x, y: cmp(x[0], y[0]))
i = 0
while i < len(intervels) - 1:
if intervels[i][1] > intervels[i + 1][0]:
intervels[i] = (intervels[i][0], intervels[i + 1][1],
intervels[i][2] and intervels[i + 1][2])
del intervels[i + 1]
else:
i = i + 1
dec = ffmpeg_decoder(mov_name, SAMPLE_RATE)
spec = spectrum(dec.ostream.get_handle(), squared=False)
feat = feature_extractor(spec.ostream.get_handle(), intervels, speech,
noise)
dec.start()
spec.start()
feat.start()
feat.join()
return speech, noise
def generate_data(sub_names, mov_names):
speech = []
noise = []
for sub_name, mov_name in zip(sub_names, mov_names):
fsub = open(sub_name)
intervels = []
while True:
if fsub.readline().strip() == '[Events]':
break
for line in fsub:
if line.startswith('Dialogue:'):
line = line.strip().split(',')
if line[3] != 'Default' or line[9].startswith('{'):
intervels.append( (parse_time(line[1]), parse_time(line[2]), False) )
else:
intervels.append( (parse_time(line[1]), parse_time(line[2]), True) )
intervels.sort(cmp=lambda x,y: cmp(x[0], y[0]))
i = 0
while i < len(intervels)-1:
if intervels[i][1] > intervels[i+1][0]:
intervels[i] = (intervels[i][0], intervels[i+1][1], intervels[i][2] and intervels[i+1][2])
del intervels[i+1]
else:
i = i + 1
dec = ffmpeg_decoder(mov_name, SAMPLE_RATE)
spec = spectrum(dec.ostream.get_handle(), squared = False)
feat = feature_extractor(spec.ostream.get_handle(), intervels, speech, noise)
dec.start()
spec.start()
feat.start()
feat.join()
return speech, noise
import core.audio_decoder as ad
import core.ffmpeg_decoder as fd
import core.sub_generator as sg
import core.spectrum as spec
from core.naive_vad2 import *
constants.DEBUG = True
dec = fd.ffmpeg_decoder('../data/demo.mp4')
vad = naive_vad(dec.ostream.get_handle())
sub = sg.sub_generator(vad.ostream.get_handle(), '../data/demo.mp4',
'../data/demo.srt')
dec.start()
vad.start()
sub.start()
sub.join()
def OnOpen(self,evt):
#self.playerpanel.OnOpen(None)
self.playerpanel.OnStop(None)
# Create a file dialog opened in the current home directory, where
# you can display all kind of files, having as title "Choose a file".
dlg = wx.FileDialog(self, "Choose a file", user.home, "","*.*", wx.OPEN)
if dlg.ShowModal() == wx.ID_OK:
dirname = dlg.GetDirectory()
filename = dlg.GetFilename()
# Creation
self.playerpanel.mediapath=unicode(os.path.join(dirname, filename))
self.playerpanel.Media = self.playerpanel.Instance.media_new(self.playerpanel.mediapath)
#m=self.Instance.media_new(r'D:\shiyan\number3\New folder\1.rmvb')
self.playerpanel.player.set_media(self.playerpanel.Media)
# Report the title of the file chosen
title = self.playerpanel.player.get_title()
# if an error was encountred while retriving the title, then use
# filename
if title == -1:
title = filename
#self.SetTitle("%s - AutoSub" % title)
# set the window id where to render VLC's video output
self.playerpanel.player.set_hwnd(self.playerpanel.videopanel.GetHandle())
# set the volume slider to the current volume
self.playerpanel.volslider.SetValue(self.playerpanel.player.audio_get_volume() / 2)
self.playerpanel.title=title
# finally destroy the dialog
dlg.Destroy()
# create the new dialog to choose the recognization and translation
self.playerpanel.select_dialog=SelectDialog(None,"Choice")
self.playerpanel.select_dialog.ShowModal()
# Finally Play~FIXME: this should be made cross-platform
self.OnPlay(None)
#self.Spec.GetAddr(self.Spec,self.mediapath)
else:
dlg.Destroy()
time.sleep(0.5)
leng=self.playerpanel.player.get_length()
self.spec.GetLength(self.spec,length=leng)
self.playerpanel.timer.Start(100)
#self.bitmap.Hide()
self.SetTitle("%s - AutoSub" % self.playerpanel.title)
lan={"English":"en" ,"Chinese":"zh-cn" ,"Japanese":"ja"}
lang_from = None
lang_to = None
source = None
target = None
if self.playerpanel.select_dialog.isrecognize==True:
# Set recognize parameter
lang_from=lan[self.playerpanel.select_dialog.sorcelan]
if self.playerpanel.select_dialog.istranslate==True:
# Set translation parameter
lang_to=lan[self.playerpanel.select_dialog.targetlan]
source=self.playerpanel.mediapath
# Set target name
if not target:
target = source[:source.rfind('.')] + '.srt'
self.subtitle=target
self.currentfile=None
self.dec = fd.ffmpeg_decoder(source,output_rate = 8000)
self.vad = naive_vad(self.dec.ostream.get_handle())
self.sub = sg.sub_generator(self.vad.ostream.get_handle(), source, target, lang_from = lang_from, lang_to = lang_to)
self.ohandle = self.sub.ostream.get_handle()
self.specd = sp.spectrum(self.dec.ostream.get_handle(), window_size = 1024)
handle = self.specd.ostream.get_handle()
#self.Spec.OpenData(self.Spec,self.ohandle)
self.dec.start()
self.vad.start()
self.sub.start()
self.specd.start()
self.spec.OpenData(self.spec,handle)
def OnOpen(self, evt):
#self.playerpanel.OnOpen(None)
self.playerpanel.OnStop(None)
# Create a file dialog opened in the current home directory, where
# you can display all kind of files, having as title "Choose a file".
dlg = wx.FileDialog(self, "Choose a file", user.home, "", "*.*",
wx.OPEN)
if dlg.ShowModal() == wx.ID_OK:
dirname = dlg.GetDirectory()
filename = dlg.GetFilename()
# Creation
self.playerpanel.mediapath = unicode(
os.path.join(dirname, filename))
self.playerpanel.Media = self.playerpanel.Instance.media_new(
self.playerpanel.mediapath)
#m=self.Instance.media_new(r'D:\shiyan\number3\New folder\1.rmvb')
self.playerpanel.player.set_media(self.playerpanel.Media)
# Report the title of the file chosen
title = self.playerpanel.player.get_title()
# if an error was encountred while retriving the title, then use
# filename
if title == -1:
title = filename
#self.SetTitle("%s - AutoSub" % title)
# set the window id where to render VLC's video output
self.playerpanel.player.set_hwnd(
self.playerpanel.videopanel.GetHandle())
# set the volume slider to the current volume
self.playerpanel.volslider.SetValue(
self.playerpanel.player.audio_get_volume() / 2)
self.playerpanel.title = title
# finally destroy the dialog
dlg.Destroy()
# create the new dialog to choose the recognization and translation
self.playerpanel.select_dialog = SelectDialog(None, "Choice")
self.playerpanel.select_dialog.ShowModal()
# Finally Play~FIXME: this should be made cross-platform
self.OnPlay(None)
#self.Spec.GetAddr(self.Spec,self.mediapath)
else:
dlg.Destroy()
time.sleep(0.5)
leng = self.playerpanel.player.get_length()
self.spec.GetLength(self.spec, length=leng)
self.playerpanel.timer.Start(100)
#self.bitmap.Hide()
self.SetTitle("%s - AutoSub" % self.playerpanel.title)
lan = {"English": "en", "Chinese": "zh-cn", "Japanese": "ja"}
lang_from = None
lang_to = None
source = None
target = None
if self.playerpanel.select_dialog.isrecognize == True:
# Set recognize parameter
lang_from = lan[self.playerpanel.select_dialog.sorcelan]
if self.playerpanel.select_dialog.istranslate == True:
# Set translation parameter
lang_to = lan[self.playerpanel.select_dialog.targetlan]
source = self.playerpanel.mediapath
# Set target name
if not target:
target = source[:source.rfind('.')] + '.srt'
self.subtitle = target
self.currentfile = None
self.dec = fd.ffmpeg_decoder(source, output_rate=8000)
self.vad = naive_vad(self.dec.ostream.get_handle())
self.sub = sg.sub_generator(self.vad.ostream.get_handle(),
source,
target,
lang_from=lang_from,
lang_to=lang_to)
self.ohandle = self.sub.ostream.get_handle()
self.specd = sp.spectrum(self.dec.ostream.get_handle(),
window_size=1024)
handle = self.specd.ostream.get_handle()
#self.Spec.OpenData(self.Spec,self.ohandle)
self.dec.start()
self.vad.start()
self.sub.start()
self.specd.start()
self.spec.OpenData(self.spec, handle)
else:
print sys.argv[i]
source = sys.argv[i]
i = i + 1
if not source:
raise ValueError()
if not target:
target = source[:source.rfind('.')] + '.srt'
except:
print 'Usage: autosub [options...] <input video>'
print 'Example: autosub -r ja -t zh-cn demo.mp4'
print 'Options:'
print ' -r <language code> enable speech recognition and set source language to <language code>'
print ' -t <language code> enable translation and set target language to <language code>'
print ' -o <output>.srt specify output subtitle file name (default: same as input)'
print 'Language codes:'
print ' Chinese zh-cn'
print ' English en'
print ' Japanese ja'
exit()
dec = fd.ffmpeg_decoder(source)
vad = naive_vad(dec.ostream.get_handle())
sub = sg.sub_generator(vad.ostream.get_handle(), source, target, lang_from = lang_from, lang_to = lang_to)
dec.start()
vad.start()
sub.start()
sub.join()
if line[3] != 'Default' or line[9].startswith('{\\an8'):
intervels.append( (parse_time(line[1]), parse_time(line[2]), False) )
else:
intervels.append( (parse_time(line[1]), parse_time(line[2]), True) )
intervels.sort(cmp=lambda x,y: cmp(x[0], y[0]))
i = 0
while i < len(intervels)-1:
if intervels[i][1] > intervels[i+1][0]:
intervels[i] = (intervels[i][0], intervels[i+1][1], intervels[i][2] and intervels[i+1][2])
del intervels[i+1]
else:
i = i + 1
dec = ffmpeg_decoder(mov_name, SAMPLE_RATE)
spec = spectrum(dec.ostream.get_handle(), squared = False)
feat = feature_extractor(spec.ostream.get_handle(), intervels, speech, noise)
dec.start()
spec.start()
feat.start()
feat.join()
random.shuffle(speech)
random.shuffle(noise)
for f in speech:
plt.imshow(np.log(abs(f)**2).reshape((5, 128)))
plt.show()
import core.audio_decoder as ad
import core.ffmpeg_decoder as fd
import core.sub_generator as sg
import core.spectrum as spec
from core.naive_vad2 import *
constants.DEBUG = True
dec = fd.ffmpeg_decoder('../data/demo.mp4')
vad = naive_vad(dec.ostream.get_handle())
sub = sg.sub_generator(vad.ostream.get_handle(), '../data/demo.mp4', '../data/demo.srt')
dec.start()
vad.start()
sub.start()
sub.join()
