def cut_video(recording_path, datapack_dir):
# Read the start/end pattern
sr1, pattern_wav = wav.read('pattern.wav')
workingdir = tempfile.mkdtemp()
# Open the video file
clip = VideoFileClip(recording_path)
# Save its audio track temporarily on disk
clip.audio.write_audiofile(os.path.join(workingdir, "temp_audio.wav"))
# Read the audio samples, mix down to mono (if necessary), and delete the temporary audio track
sr2, recording_wav = wav.read(os.path.join(workingdir, "temp_audio.wav"))
if recording_wav.shape[1] > 1:
recording_wav = numpy.mean(recording_wav, 1)
shutil.rmtree(workingdir)
# Detect the start and end audio pattern
start, end = detect_start_end_times(pattern_wav, recording_wav, sr2, 4)
# Cut the video and write it into two separate video and audio files
clip.subclip(start + 0.4,
end).write_videofile(os.path.join(datapack_dir, 'video.mp4'),
codec='libx264')
clip.subclip(start + 0.4, end).audio.write_audiofile(
os.path.join(datapack_dir, 'audio.wav'))
def cut_video(recording_path, datapack_dir):
# Read the start/end pattern
sr1, pattern_wav = wav.read('pattern.wav')
workingdir = tempfile.mkdtemp()
# Open the video file
clip = VideoFileClip(recording_path)
# Save its audio track temporarily on disk
clip.audio.write_audiofile(os.path.join(workingdir,"temp_audio.wav"))
# Read the audio samples, mix down to mono (if necessary), and delete the temporary audio track
sr2, recording_wav = wav.read(os.path.join(workingdir,"temp_audio.wav"))
if recording_wav.shape[1]>1:
recording_wav = numpy.mean(recording_wav,1)
shutil.rmtree(workingdir)
# Detect the start and end audio pattern
start, end = detect_start_end_times(pattern_wav, recording_wav, sr2, 4)
# Cut the video and write it into two separate video and audio files
clip.subclip(start+0.4, end).write_videofile(os.path.join(datapack_dir, 'video.mp4'), codec='libx264')
clip.subclip(start+0.4, end).audio.write_audiofile(os.path.join(datapack_dir,'audio.wav'))
def cutVideo(filepaths):
############################################ Initializing ########################################################
for filepath in filepaths:
video = VideoFileClip(filepath)
duration = int(video.duration)
size = int(duration / 15)
real_size = duration / 15
if (size > real_size):
size = size - 1
end_custom = (15 * size) + (duration % 15)
start_time = 0
end_time = 15
j = 0
################################## Check if Folder Already Exists ###############################################
for j in range(100):
if (os.path.isdir(os.getcwd() + '/Stories' + '0' +
str(j + 1)) == False):
os.mkdir(os.getcwd() + '/Stories' + '0' + str(j + 1))
j += 1
break
path = (os.getcwd() + '/Stories' + '0' + str(j))
################################## If duration is divisible by 15 ###############################################
if (duration % 15 == 0):
for i in range(size):
#if (abort == None):
if (end_time > duration + (i - 1 * 0.01)):
break
else:
new = video.subclip(start_time, end_time)
new.write_videofile(path + "/stories" + '_0' + str(i + 1) +
".mp4",
threads=8,
preset="veryfast",
audio_codec='aac')
start_time += 15.01
end_time += 15.01
################################## If duration is NOT divisible by 15 ###############################################
else:
i = 0
while i < size:
new = video.subclip(start_time, end_time)
new.write_videofile(path + "/stories" + '_0' + str(i + 1) +
".mp4",
threads=8,
preset="veryfast",
audio_codec='aac')
start_time += 15.01
end_time += 15.01
i += 1
if (i == size):
new = video.subclip(start_time + 0.01, end_custom)
new.write_videofile(path + "/stories" + '_0' + str(size + 1) +
".mp4",
threads=8,
preset="veryfast",
audio_codec='aac')
def clip(args):
'''Subprogram for taking a portion of a video'''
help(
'{script} [--help] [--threads THREADS] [--output OUTPUT_FILE] [--mute] INPUT_FILE START_TIME END_TIME'
)
value_flags = ['-o', '--output', '-t', '--threads']
input_file = unnamed_value(0, value_flags, argl=args)
start_time = unnamed_value(1, value_flags, argl=args)
end_time = unnamed_value(2, value_flags, argl=args)
mute = has_any(['-m', '--mute'])
output_file = value_following_any(['-o', '--output'], argl=args)
if output_file is None:
_, ext = splitext(input_file)
output_file = generate_filename(ext)
threads = 8
if has_any(['-t', '--threads'], argl=args):
threads = value_following_any(['-t', '--threads'], argl=args)
clip = VideoFileClip(input_file)
subclip = clip.subclip(start_time, end_time)
subclip.write_videofile(output_file, threads=threads, audio=not mute)
clip.close()
subclip.close()
def delmp4(filename,time1=0,time2=0,mv=0):
#剪切视频文件 路径文件名 去片头 去片尾 到片尾时间
clip = VideoFileClip(filename)
clip_len=clip.duration
if time2==0 and mv>0:
mvtime=mv
else:
mvtime=int(clip_len-time2)
if (clip_len-time1-time2)>0 and time1<clip_len and time2<clip_len and mv<clip_len:
#if ((clip_len-time1-time2)>0 and mvtime==mv) or (mv>time1 and mv<clip_len and mv>time2) :
mvclip=clip.subclip(time1,mvtime)
file_name = os.path.splitext(filename)[0]
filen= os.path.basename(filename)
mvclip.write_videofile(work_path+"\\"+filen)
mvclip.close()
#防止内存溢出,用完关闭!!!
else:
print("视频长度不够!!!!")
if not os.path.isdir(work_path):
os.makedirs(work_path)
fo = open(work_path+"/错误!视频长度不够.txt", "a+")
fo.write( filename+"\n")
fo.close()
#shutil.copy(file_name,len_path)
clip.close()
def summarize(filepath, new_filename, hotclips):
"""
Inputs a filepath for a video and generates a new shorter video
in that same filepath.
"""
# Only open the file once!
video = VideoFileClip(filepath)
chunks = [video.subclip(start, end)
for (start, end) in hotclips]
final_clip = concatenate(chunks)
# txt_clip = ( TextClip("Generated by vSummarize",
# fontsize=20, color='white')
# .set_pos('bottom')
# .set_duration(5))
# final_clip = CompositeVideoClip([summarized_video, txt_clip])
# Use the to_videofile default codec, libx264
# libx264 is much better than mpeg4, and still writes .mp4
# Use the fps of the original video.
final_clip.to_videofile(new_filename,
fps=video.fps,
audio_codec='mp3')
def process(self):
'''
Process video
'''
video_clip = VideoFileClip(self.video_path)
if self.start or self.end:
video_clip = video_clip.subclip(self.start, self.end)
video_clip = video_clip.set_fps(video_clip.fps // 10)
total_frames = round(video_clip.fps * video_clip.duration)
print(
f'Processing video {self.video_path}, {total_frames} frames, size: {video_clip.size}'
)
for frame in tqdm(video_clip.iter_frames()):
# It's better if we resizing before crop to keep the image looks more 'scene'
h, w, _ = frame.shape
aspect_ratio = w / h
ratio = h / (self.image_size * 2)
w /= ratio
h = w / aspect_ratio
frame = cv2.resize(frame[..., ::-1], (int(h), int(w)))
if self.crop_and_save(frame):
break
print(f'Saved {self.counter} images to {self.save_dir}')
def getVCItem(clipIndex, meanPreCutLen):
vc = VideoFileClip("movie/clip" + str(clipIndex) + ".mp4")
vcLen = vc.duration
startPoint = vcLen / 2 - meanPreCutLen
endPont = vcLen / 2
vcItem = vc.subclip(int(startPoint), int(endPont))
return vcItem
def main(source):
print('要剪切的视频是', source[1]) # 获取需要剪切的文件
source = source[1]
time = input('请输入开始和结束的时间:\n') # 获取开始剪切时间
time = time.split()
times = []
while time != []:
start_time = time.pop(0)
stop_time = time.pop(0)
times.append([start_time, stop_time])
print("子视频命名方式为原名称_段数.mp4")
temp = source.split('.')
target = []
num_of_cuts = len(times)
for i in list(range(1, num_of_cuts + 1)):
target.append(temp[0] + str(i) + '.' + temp[1])
# print('**--**',target)
video = VideoFileClip(source)
for i in list(range(0, num_of_cuts, 1)):
current_time = times[i]
print('cutting...\n')
temp = list(map(int, current_time[0].split(":")))
start_time = reduce(to_seconds, temp)
temp = list(map(int, current_time[1].split(":")))
stop_time = reduce(to_seconds, temp)
new_video = video.subclip(start_time, stop_time) # 执行剪切操作
new_video.to_videofile(target[i], fps=29, remove_temp=True) # 输出文件
video.close()
print('done ^_^\n')
def generateAudioClips(clip: VideoFileClip, segments: List[Segment],
output_dir):
# this needs to be done in two loops to avoid assertions from the sub clipping
for seg in segments:
seg.video = clip.subclip(seg.startTime, seg.endTime)
for seg in segments:
seg.audioPath = output_dir + "/clip_" + str(seg.startTime) + ".mp3"
seg.video.audio.write_audiofile(seg.audioPath)
def video_frames(path: Path, t_start=0, t_end=None) -> Iterator[np.ndarray]:
"""Video frames iterator that releases the resources correctly."""
clip = VideoFileClip(str(path))
try:
subclip = clip.subclip(t_start=t_start, t_end=t_end)
yield from subclip.iter_frames()
finally:
## https://stackoverflow.com/questions/43966523/getting-oserror-winerror-6-the-handle-is-invalid-in-videofileclip-function
clip.reader.close()
clip.audio.reader.close_proc()
def cut_one(self,row):
start = self.get_seconds_start(row)
end = start + row['toAdd']
self.video_id = f'_{start}{end}.mp4'
name = f'{self.highlights_location}/video{self.video_id}'
fajl = VideoFileClip(self.mc["src"])
new = fajl.subclip(start,end)
new.write_videofile(name,logger= None)
progress = self.progress_part * (self.highlights.index(row)+1)
self.update_cut_progress(progress)
def addNum(self):
source = self.source_le.text().strip() #获取需要剪切的文件
target = self.target_le.text().strip() #获取剪切后视频保存的文件
start_time = self.start_le.text().strip() #获取开始剪切时间
stop_time = self.stop_le.text().strip() #获取剪切的结束时间
video = VideoFileClip(source) #视频文件加载
video = video.subclip(int(start_time), int(stop_time)) #执行剪切操作
video.to_videofile(target, fps=20, remove_temp=True) #输出文件
self.result_le.setText("ok!") #输出文件后界面返回OK
self.result_le.setStyleSheet(
"color:red;font-size:40px") #设置OK颜色为红色，大小为四十像素
self.result_le.setAlignment(Qt.AlignCenter) #OK在指定框内居中
def clip(args: Namespace):
'''Subprogram for taking a portion of a video'''
output_file = args.output
if output_file is None:
_, ext = splitext(args.input_file)
print(f'Output file using same extension as input file: {ext}')
output_file = generate_filename(ext)
audio = not args.mute
clip = VideoFileClip(args.input_file)
subclip = clip.subclip(args.start_time, args.end_time)
subclip.write_videofile(output_file, threads=args.threads, audio=audio)
clip.close()
subclip.close()
def getIntroVid(vidTitle, origCropFrac, sampleHeight, origVidDir, username):
if len(username)==0:
username="******"
titleParts = [vidTitle,
" ",
"by \n " + username]
sizes = [50,
30,
80]
introVid = VideoFileClip(origVidDir)
(w, h) = introVid.size
length = introVid.duration
times = [[0, int(length/3)],
[int(length/3), int(length*2/3)],
[int(length*2/3), int(length)]]
# times = [[0, 3],
# [3, 9],
# [9, 14]]
introVid = introVid.crop(height=int(round((1 - origCropFrac*2) * sampleHeight, 0)), width = w, x_center=w/2, y_center=h/2)
clips=[]
iPart=0
while iPart < len(titleParts):
title=titleParts[iPart]
if times[iPart][1]>introVid.duration:
times[iPart][1] = introVid.duration
vidClip = introVid.subclip(times[iPart][0], times[iPart][1])
text = TextClip(title, font="Amiri-Bold", fontsize=sizes[iPart], color="white", align = 'center').set_position(("center",0.3), relative=True)
text_clip = mpe.CompositeVideoClip([vidClip, text]).set_duration(vidClip.duration)
clips.append(text_clip)
iPart = iPart + 1
final_clip = concatenate_videoclips(clips, method='compose')
final_clip1 = fadeout(final_clip, 1, final_color=None)
final_clip2 = fadein(final_clip1, 1, initial_color=None)
return final_clip2
def test_FramesMatches_select_scenes(
filename,
subclip,
match_threshold,
min_time_span,
nomatch_threshold,
expected_result,
):
video_clip = VideoFileClip(filename)
if subclip is not None:
video_clip = video_clip.subclip(subclip[0], subclip[1])
clip = concatenate_videoclips([video_clip.fx(time_mirror), video_clip])
result = FramesMatches.from_clip(clip, 10, 3, logger=None).select_scenes(
match_threshold,
min_time_span,
nomatch_threshold=nomatch_threshold,
)
assert len(result) == len(expected_result)
assert result == expected_result
def crop_word(found_word: str, start: float, end: float, videoid: str,
save_loc: Tuple[str]):
# pylint: disable=too-many-arguments, too-many-locals
'''crops a video into a small segment, including padding, face detection
and face bounding.'''
# get full video
full_video = VideoFileClip(f'videos/{videoid}.mp4')
# pad video
pad = 0.25
start, end = apply_padding(start, end, pad, full_video.duration)
# get subclip
word_subclip = full_video.subclip(start, end)
# get frame
start_frame = word_subclip.get_frame(t=0)
end_frame = word_subclip.get_frame(t=end - start)
# detect faces
start_faces = get_faces(start_frame)
end_faces = get_faces(end_frame)
if len(start_faces) == 1 and len(end_faces) == 1:
bound_face = get_face_bounds(start_faces[0], end_faces[0])
final_word = crop(
word_subclip,
x1=bound_face['left'],
x2=bound_face['right'],
y1=bound_face['top'],
y2=bound_face['bottom'],
)
filename = f'{videoid}_{start:.2f}'
save_to_file(final_word, found_word, filename, save_loc)
elif len(start_faces) == 0 or len(end_faces) == 0:
print('No faces found in either the start or end frame.')
else:
print('Multiple faces found')
def summarize(filepath, new_filename, hotclips):
"""
Inputs a filepath for a video and generates a new shorter video
in that same filepath.
"""
# Only open the file once!
video = VideoFileClip(filepath)
chunks = [video.subclip(start, end) for (start, end) in hotclips]
final_clip = concatenate(chunks)
# txt_clip = ( TextClip("Generated by vSummarize",
# fontsize=20, color='white')
# .set_pos('bottom')
# .set_duration(5))
# final_clip = CompositeVideoClip([summarized_video, txt_clip])
# Use the to_videofile default codec, libx264
# libx264 is much better than mpeg4, and still writes .mp4
# Use the fps of the original video.
final_clip.to_videofile(new_filename, fps=video.fps, audio_codec='mp3')
def video_clip(path, time_start=0, clip_seconds=10, unique_id=None):
assert isinstance(time_start, int) and time_start >= 0
assert isinstance(clip_seconds, int) and clip_seconds > 0
clip = VideoFileClip(path)
time_end = time_start + clip_seconds
if time_start > clip.duration:
time_start = clip.duration
if time_end > clip.duration:
time_end = clip.duration
if time_start - time_end == 0:
raise core_exceptions.ZeroLengthError()
if unique_id is None:
path = os.path.abspath(path)
unique_id = FNV64.hash(path)
output_name = f"{unique_id}_{time_start}_{clip_seconds}.mp4"
print("Clip from", time_start, "to", time_end, "sec in:", output_name)
sub_clip = clip.subclip(time_start, time_end)
sub_clip.write_videofile(output_name)
sub_clip.close()
clip.close()
del clip
del sub_clip
return output_name
def summarize(self):
print('Fetching features...')
self.__fetch_features()
print('Features fetched...')
self.__get_n_frames()
print('Getting `n` frames...')
self.__set_scene_detector()
print('Scene detector set; detecting scenes ...')
self.__detect_scenes()
print('Scenes detected, selecting scenes ...')
selected_scenes = self.__knapsack()
print('Scenes selected, generating summary ...')
original = VideoFileClip(self.F.video_filepath)
clips = []
for i in range(1, len(selected_scenes)):
start = self.__timecodes[selected_scenes[i - 1]]
end = self.__timecodes[selected_scenes[i - 1] + 1]
clip = original.subclip(start, end)
clips.append(clip)
summ = concatenate_videoclips(clips)
summ.write_videofile(self.__outpath)
return self.__outpath, "Success"
def trim_silences(self, silences: List[Tuple[float, float, float]]):
video = VideoFileClip(self.in_file)
full_duration = video.duration
# print(f"pre_silences: {silences}")
_silences = self.rescale_all_silences(silences, min_duration=self.min_duration, max_duration=self.max_duration)
# print(f"post_silences: {_silences}")
non_silent_periods = list(
[(end1, start2 - end1, start2) for (_, _, end1), (start2, _, _) in zip(_silences[:-1], _silences[1:])])
print(non_silent_periods)
input_dir, input_file = os.path.split(self.in_file)
fname, fext = os.path.splitext(input_file)
output_fname = os.path.join(input_dir, f"{fname}_NOSILENCE_{self.min_duration}s{fext}")
tmp_audio_fname = f"{fname}.TEMP_MPY_wvf_snd_custom.ogg"
tmp_video_fname = f"{fname}.TEMP_MPY_vid_custom{fext}"
print(f"writing output to {output_fname}")
clips = list([video.subclip(s, e) for s, d, e in non_silent_periods if d >= 1])
print(f"got list of clips")
# comp = mpy.CompositeVideoClip(clips)
comp = concatenate_videoclips(clips)
print(f"make composite video clip (no sound yet)")
comp.write_videofile(tmp_video_fname, codec=self.codec, preset='ultrafast',
threads=os.cpu_count() + 1, audio_codec=self.audio_codec)
print(f"done writing out ${tmp_video_fname}")
video.close()
print(f"closed video")
# print(f"preparing to write audio")
# comp.audio.write_audiofile(tmp_audio_fname, buffersize=2000000, codec=self.audio_codec)
# print(f"wrote audio")
# comp.audio = None
# print(f"wrote out video, now combining")
# output_video = VideoFileClip(tmp_video_fname, audio=False)
# output_video.set_audio(AudioFileClip(tmp_audio_fname))
# output_video.write_videofile(output_fname)
# output_video.write_videofile(output_fname, preset='ultrafast', codec=self.codec, threads=os.cpu_count() + 1,
# fps=video.fps)
print(f"wrote video out")
def process_video(self,
video_file,
file_out,
t_start=None,
t_end=None,
process_pool=None):
input_clip = VideoFileClip(video_file)
if t_start is not None:
input_clip = input_clip.subclip(t_start=t_start, t_end=t_end)
if self.debug:
self.processed_frames = []
stage_idx = 0
output_clip = input_clip.fl_image(
lambda frame: self.process_frame_stage(frame, stage_idx,
process_pool))
output_clip.write_videofile(file_out, audio=False)
if len(self.processed_frames) > 0:
out_file_path = os.path.split(file_out)
out_file_name = out_file_path[1].split('.')
for _ in range(len(self.processed_frames[0]) - 1):
self.frame_count = 0
stage_idx += 1
stage_file = '{}.{}'.format(
os.path.join(out_file_path[0], out_file_name[0]) +
'_' + str(stage_idx), out_file_name[1])
output_clip.write_videofile(stage_file, audio=False)
else:
output_clip = input_clip.fl_image(
lambda frame: self.process_frame(frame, process_pool))
output_clip.write_videofile(file_out, audio=False)
def instaTV(url, firefox, button, bot, chat_id):
bot.send_message(chat_id=chat_id,
text='Isso pode demorar um pouco',
timeout=60)
temp_dir = tempfile.mkdtemp()
if (auth(url, firefox, button, bot) == True):
firefox.get(url)
time.sleep(1)
html = firefox.page_source
time.sleep(1)
encodedVideo = re.search(r'"video_url":"(.*?)",', html)
if (encodedVideo == None):
decoded = ''
return decoded
else:
if (encodedVideo != None):
path = temp_dir + '/IGTV.mp4'
decoded = (encodedVideo.group(1).replace(r"\u0026", "&"))
urllib.request.urlretrieve(decoded, path)
size = os.path.getsize(path) / 1000000
if size <= 49.9999:
bot.send_video(chat_id=chat_id,
video=open(path, 'rb'),
timeout=3600)
shutil.rmtree(temp_dir)
return path
else:
video = VideoFileClip(path)
duration = int(video.duration)
start_time = 0
end_time = 600
parts = int(duration / 600)
videopart = int(duration / 600) + 1
end_custom = (600 * parts) + (duration % 600)
igtvPath = []
if (parts > (video.duration / 600)):
parts = parts - 1
if (duration <= 630):
bot.send_message(chat_id=chat_id,
text='Finalizando ' + runner,
timeout=60)
subvideo = video.subclip(0, duration)
subvideo.write_videofile(temp_dir + '/IGTV' + '01' +
'.mp4',
threads=8,
preset="veryfast",
logger=None,
audio_codec='aac',
rewrite_audio=False,
fps=30,
bitrate='520k')
time.sleep(1)
bot.send_video(chat_id=chat_id,
video=open(
temp_dir + '/IGTV' + '01' + '.mp4',
'rb'),
timeout=3600)
igtvPath = temp_dir + '/IGTV01.mp4'
shutil.rmtree(temp_dir)
return igtvPath
else:
if (duration % 600 == 0):
c = 1
for i in range(parts):
if (end_time > duration):
break
subvideo = video.subclip(start_time, end_time)
bot.send_message(chat_id=chat_id,
text='Processando ' +
'parte ' + str(c) + '/' +
str(videopart) + '' + load,
timeout=60)
subvideo.write_videofile(temp_dir + '/IGTV' +
'0' + str(i + 1) +
'.mp4',
threads=8,
preset="veryfast",
logger=None,
audio_codec='aac',
rewrite_audio=False,
fps=30,
bitrate='550k')
time.sleep(1)
bot.send_video(chat_id=chat_id,
video=open(
temp_dir + '/IGTV' + '0' +
str(i + 1) + '.mp4', 'rb'),
timeout=3600)
c += 1
start_time += 600.00
end_time += 600.00
elif (duration % 600 > 0):
i = 0
c = 1
while i < parts:
subvideo = video.subclip(start_time, end_time)
bot.send_message(chat_id=chat_id,
text='Processando ' +
'parte ' + str(c) + '/' +
str(videopart) + '' + load,
timeout=60)
subvideo.write_videofile(temp_dir + '/IGTV' +
'0' + str(i + 1) +
'.mp4',
threads=8,
preset="veryfast",
logger=None,
audio_codec='aac',
rewrite_audio=False,
fps=30,
bitrate='550k')
time.sleep(1)
bot.send_video(chat_id=chat_id,
video=open(
temp_dir + '/IGTV' + '0' +
str(i + 1) + '.mp4', 'rb'),
timeout=3600)
c += 1
start_time += 600.00
end_time += 600.00
i += 1
if (i == parts):
subvideo = video.subclip(
start_time, end_custom)
bot.send_message(chat_id=chat_id,
text='Processando ' +
'parte ' + str(c) + '/' +
str(videopart) + '' + load,
timeout=60)
subvideo.write_videofile(temp_dir + '/IGTV' +
'0' + str(i + 1) +
'.mp4',
threads=8,
preset="veryfast",
logger=None,
audio_codec='aac',
rewrite_audio=False,
fps=30,
bitrate='550k')
time.sleep(1)
bot.send_video(chat_id=chat_id,
video=open(
temp_dir + '/IGTV' + '0' +
str(i + 1) + '.mp4', 'rb'),
timeout=3600)
igtvPath = temp_dir
shutil.rmtree(temp_dir)
return igtvPath
else:
decoded = ''
return decoded
# Hit 'q' on the keyboard to quit!
if cv2.waitKey(1) & 0xFF == ord('q'):
video_capture.release()
break
cv2.destroyAllWindows()
exit()
media_path = self._video_path({ 'name' : video_file })
if not photos:
# Process video; start loading the video clip
video = VideoFileClip(media_path)
# If a duration is set, trim clip
if duration:
video = video.subclip(start_time, start_time + duration)
# Resize clip before processing
if width:
video = video.resize(width = width)
# Crop clip if desired
if crop_x:
video = video.fx(crop, x2 = video.w / 2)
# Kick off convert frames for each frame
new_video = video.fl(_convert_helper)
# Stack clips side by side
if side_by_side:
def add_caption(caption, clip):
class MovieStim3(BaseVisualStim, ContainerMixin):
"""A stimulus class for playing movies (mpeg, avi, etc...) in PsychoPy
that does not require avbin. Instead it requires the cv2 python package
for OpenCV. The VLC media player also needs to be installed on the
psychopy computer.
**Example**::
See Movie2Stim.py for demo.
"""
def __init__(self, win,
filename="",
units='pix',
size=None,
pos=(0.0, 0.0),
ori=0.0,
flipVert=False,
flipHoriz=False,
color=(1.0, 1.0, 1.0),
colorSpace='rgb',
opacity=1.0,
volume=1.0,
name='',
loop=False,
autoLog=True,
depth=0.0,
noAudio=False,
vframe_callback=None,
fps=None,
interpolate=True):
"""
:Parameters:
filename :
a string giving the relative or absolute path to the movie.
flipVert : True or *False*
If True then the movie will be top-bottom flipped
flipHoriz : True or *False*
If True then the movie will be right-left flipped
volume :
The nominal level is 100, and 0 is silence.
loop : bool, optional
Whether to start the movie over from the beginning if draw is
called and the movie is done.
"""
# what local vars are defined (these are the init params) for use
# by __repr__
self._initParams = dir()
self._initParams.remove('self')
super(MovieStim3, self).__init__(win, units=units, name=name,
autoLog=False)
retraceRate = win._monitorFrameRate
if retraceRate is None:
retraceRate = win.getActualFrameRate()
if retraceRate is None:
logging.warning("FrameRate could not be supplied by psychopy; "
"defaulting to 60.0")
retraceRate = 60.0
self._retraceInterval = 1.0 / retraceRate
self.filename = filename
self.loop = loop
self.flipVert = flipVert
self.flipHoriz = flipHoriz
self.pos = numpy.asarray(pos, float)
self.depth = depth
self.opacity = float(opacity)
self.interpolate = interpolate
self.noAudio = noAudio
self._audioStream = None
self.useTexSubImage2D = True
self._videoClock = Clock()
self.loadMovie(self.filename)
self.setVolume(volume)
self.nDroppedFrames = 0
# size
if size is None:
self.size = numpy.array([self._mov.w, self._mov.h],
float)
else:
self.size = val2array(size)
self.ori = ori
self._updateVertices()
# set autoLog (now that params have been initialised)
self.autoLog = autoLog
if autoLog:
logging.exp("Created %s = %s" % (self.name, str(self)))
def reset(self):
self._numpyFrame = None
self._nextFrameT = None
self._texID = None
self.status = NOT_STARTED
def setMovie(self, filename, log=True):
"""See `~MovieStim.loadMovie` (the functions are identical).
This form is provided for syntactic consistency with other visual
stimuli.
"""
self.loadMovie(filename, log=log)
def loadMovie(self, filename, log=True):
"""Load a movie from file
:Parameters:
filename: string
The name of the file, including path if necessary
After the file is loaded MovieStim.duration is updated with the movie
duration (in seconds).
"""
self.reset() # set status and timestamps etc
# Create Video Stream stuff
if os.path.isfile(filename):
self._mov = VideoFileClip(filename, audio=(1 - self.noAudio))
if (not self.noAudio) and (self._mov.audio is not None):
try:
self._audioStream = sound.Sound(
self._mov.audio.to_soundarray(),
sampleRate=self._mov.audio.fps)
except:
# JWE added this as a patch for a moviepy oddity where the duration is inflated in the saved file
# causes the audioclip to be the wrong length, so round down and it should work
jwe_tmp = self._mov.subclip(0,round(self._mov.duration))
self._audioStream = sound.Sound(
jwe_tmp.audio.to_soundarray(),
sampleRate=self._mov.audio.fps)
del(jwe_tmp)
else: # make sure we set to None (in case prev clip had audio)
self._audioStream = None
else:
raise IOError("Movie file '%s' was not found" % filename)
# mov has attributes:
# size, duration, fps
# mov.audio has attributes
# duration, fps (aka sampleRate), to_soundarray()
self._frameInterval = 1.0 / self._mov.fps
self.duration = self._mov.duration
self.filename = filename
self._updateFrameTexture()
logAttrib(self, log, 'movie', filename)
def play(self, log=True):
"""Continue a paused movie from current position.
"""
status = self.status
if self._audioStream is not None:
self._audioStream.play()
if status != PLAYING:
self.status = PLAYING
self._videoClock.reset(-self.getCurrentFrameTime())
if status == PAUSED:
self._audioSeek(self.getCurrentFrameTime())
if log and self.autoLog:
self.win.logOnFlip("Set %s playing" % (self.name),
level=logging.EXP, obj=self)
self._updateFrameTexture()
def pause(self, log=True):
"""
Pause the current point in the movie (sound will stop, current frame
will not advance). If play() is called again both will restart.
"""
if self.status == PLAYING:
self.status = PAUSED
if self._audioStream:
self._audioStream.stop()
if log and self.autoLog:
self.win.logOnFlip("Set %s paused" %
(self.name), level=logging.EXP, obj=self)
return True
if log and self.autoLog:
self.win.logOnFlip("Failed Set %s paused" %
(self.name), level=logging.EXP, obj=self)
return False
def stop(self, log=True):
"""Stop the current point in the movie (sound will stop, current frame
will not advance). Once stopped the movie cannot be restarted -
it must be loaded again. Use pause() if you may need to restart
the movie.
"""
if self.status != STOPPED:
self.status = STOPPED
self._unload()
self._reset()
if log and self.autoLog:
self.win.logOnFlip("Set %s stopped" % (self.name),
level=logging.EXP, obj=self)
def setVolume(self, volume):
pass # to do
def setFlipHoriz(self, newVal=True, log=True):
"""If set to True then the movie will be flipped horizontally
(left-to-right). Note that this is relative to the original,
not relative to the current state.
"""
self.flipHoriz = newVal
logAttrib(self, log, 'flipHoriz')
self._needVertexUpdate = True
def setFlipVert(self, newVal=True, log=True):
"""If set to True then the movie will be flipped vertically
(top-to-bottom). Note that this is relative to the original,
not relative to the current state.
"""
self.flipVert = newVal
logAttrib(self, log, 'flipVert')
self._needVertexUpdate = True
def getFPS(self):
"""
Returns the movie frames per second playback speed.
"""
return self._mov.fps
def getCurrentFrameTime(self):
"""Get the time that the movie file specified the current
video frame as having.
"""
return self._nextFrameT - self._frameInterval
def _updateFrameTexture(self):
if self._nextFrameT is None:
# movie has no current position, need to reset the clock
# to zero in order to have the timing logic work
# otherwise the video stream would skip frames until the
# time since creating the movie object has passed
self._videoClock.reset()
self._nextFrameT = 0
# only advance if next frame (half of next retrace rate)
if self._nextFrameT > self.duration:
self._onEos()
elif self._numpyFrame is not None:
if self._nextFrameT > (self._videoClock.getTime() -
self._retraceInterval / 2.0):
return None
self._numpyFrame = self._mov.get_frame(self._nextFrameT)
useSubTex = self.useTexSubImage2D
if self._texID is None:
self._texID = GL.GLuint()
GL.glGenTextures(1, ctypes.byref(self._texID))
useSubTex = False
# bind the texture in openGL
GL.glEnable(GL.GL_TEXTURE_2D)
# bind that name to the target
GL.glBindTexture(GL.GL_TEXTURE_2D, self._texID)
# makes the texture map wrap (this is actually default anyway)
GL.glTexParameteri(
GL.GL_TEXTURE_2D, GL.GL_TEXTURE_WRAP_S, GL.GL_REPEAT)
# data from PIL/numpy is packed, but default for GL is 4 bytes
GL.glPixelStorei(GL.GL_UNPACK_ALIGNMENT, 1)
# important if using bits++ because GL_LINEAR
# sometimes extrapolates to pixel vals outside range
if self.interpolate:
GL.glTexParameteri(
GL.GL_TEXTURE_2D, GL.GL_TEXTURE_MAG_FILTER, GL.GL_LINEAR)
GL.glTexParameteri(
GL.GL_TEXTURE_2D, GL.GL_TEXTURE_MIN_FILTER, GL.GL_LINEAR)
if useSubTex is False:
GL.glTexImage2D(GL.GL_TEXTURE_2D, 0, GL.GL_RGB8,
self._numpyFrame.shape[1],
self._numpyFrame.shape[0], 0,
GL.GL_RGB, GL.GL_UNSIGNED_BYTE,
self._numpyFrame.ctypes)
else:
GL.glTexSubImage2D(GL.GL_TEXTURE_2D, 0, 0, 0,
self._numpyFrame.shape[1],
self._numpyFrame.shape[0],
GL.GL_RGB, GL.GL_UNSIGNED_BYTE,
self._numpyFrame.ctypes)
else:
GL.glTexParameteri(
GL.GL_TEXTURE_2D, GL.GL_TEXTURE_MAG_FILTER, GL.GL_NEAREST)
GL.glTexParameteri(
GL.GL_TEXTURE_2D, GL.GL_TEXTURE_MIN_FILTER, GL.GL_NEAREST)
if useSubTex is False:
GL.glTexImage2D(GL.GL_TEXTURE_2D, 0, GL.GL_RGB8,
self._numpyFrame.shape[1],
self._numpyFrame.shape[0], 0,
GL.GL_BGR, GL.GL_UNSIGNED_BYTE,
self._numpyFrame.ctypes)
else:
GL.glTexSubImage2D(GL.GL_TEXTURE_2D, 0, 0, 0,
self._numpyFrame.shape[1],
self._numpyFrame.shape[0],
GL.GL_BGR, GL.GL_UNSIGNED_BYTE,
self._numpyFrame.ctypes)
GL.glTexEnvi(GL.GL_TEXTURE_ENV, GL.GL_TEXTURE_ENV_MODE,
GL.GL_MODULATE) # ?? do we need this - think not!
if not self.status == PAUSED:
self._nextFrameT += self._frameInterval
def draw(self, win=None):
"""Draw the current frame to a particular visual.Window (or to the
default win for this object if not specified). The current
position in the movie will be determined automatically.
This method should be called on every frame that the movie is
meant to appear.
"""
if (self.status == NOT_STARTED or
(self.status == FINISHED and self.loop)):
self.play()
elif self.status == FINISHED and not self.loop:
return
if win is None:
win = self.win
self._selectWindow(win)
self._updateFrameTexture() # will check if it's needed
# scale the drawing frame and get to centre of field
GL.glPushMatrix() # push before drawing, pop after
# push the data for client attributes
GL.glPushClientAttrib(GL.GL_CLIENT_ALL_ATTRIB_BITS)
self.win.setScale('pix')
# move to centre of stimulus and rotate
vertsPix = self.verticesPix
# bind textures
GL.glActiveTexture(GL.GL_TEXTURE1)
GL.glBindTexture(GL.GL_TEXTURE_2D, 0)
GL.glEnable(GL.GL_TEXTURE_2D)
GL.glActiveTexture(GL.GL_TEXTURE0)
GL.glBindTexture(GL.GL_TEXTURE_2D, self._texID)
GL.glEnable(GL.GL_TEXTURE_2D)
# sets opacity (1,1,1 = RGB placeholder)
GL.glColor4f(1, 1, 1, self.opacity)
array = (GL.GLfloat * 32)(
1, 1, # texture coords
vertsPix[0, 0], vertsPix[0, 1], 0., # vertex
0, 1,
vertsPix[1, 0], vertsPix[1, 1], 0.,
0, 0,
vertsPix[2, 0], vertsPix[2, 1], 0.,
1, 0,
vertsPix[3, 0], vertsPix[3, 1], 0.,
)
# 2D texture array, 3D vertex array
GL.glInterleavedArrays(GL.GL_T2F_V3F, 0, array)
GL.glDrawArrays(GL.GL_QUADS, 0, 4)
GL.glPopClientAttrib()
GL.glPopAttrib()
GL.glPopMatrix()
# unbind the textures
GL.glActiveTexture(GL.GL_TEXTURE0)
GL.glBindTexture(GL.GL_TEXTURE_2D, 0)
GL.glEnable(GL.GL_TEXTURE_2D) # implicitly disables 1D
def seek(self, t):
"""Go to a specific point in time for both the audio and video streams
"""
# video is easy: set both times to zero and update the frame texture
self._nextFrameT = t
self._videoClock.reset(t)
self._audioSeek(t)
def _audioSeek(self, t):
# for sound we need to extract the array again and just begin at new
# loc
if self._audioStream is None:
return # do nothing
self._audioStream.stop()
sndArray = self._mov.audio.to_soundarray()
startIndex = int(t * self._mov.audio.fps)
self._audioStream = sound.Sound(
sndArray[startIndex:, :], sampleRate=self._mov.audio.fps)
self._audioStream.play()
def _getAudioStreamTime(self):
return self._audio_stream_clock.getTime()
def _unload(self):
try:
# remove textures from graphics card to prevent crash
self.clearTextures()
except Exception:
pass
self._mov = None
self._numpyFrame = None
self._audioStream = None
self.status = FINISHED
def _onEos(self):
if self.loop:
self.seek(0.0)
else:
self.status = FINISHED
self.stop()
if self.autoLog:
self.win.logOnFlip("Set %s finished" % self.name,
level=logging.EXP, obj=self)
def __del__(self):
self._unload()
def setAutoDraw(self, val, log=None):
"""Add or remove a stimulus from the list of stimuli that will be
automatically drawn on each flip
:parameters:
- val: True/False
True to add the stimulus to the draw list, False to remove it
"""
if val:
self.play(log=False) # set to play in case stopped
else:
self.pause(log=False)
# add to drawing list and update status
setAttribute(self, 'autoDraw', val, log)
def convert(self,
video_file,
swap_model=False,
duration=None,
start_time=None,
use_gan=False,
face_filter=False,
photos=True,
crop_x=None,
width=None,
side_by_side=False,
live=False):
# Magic incantation to not have tensorflow blow up with an out of memory error.
import tensorflow as tf
import keras.backend.tensorflow_backend as K
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
config.gpu_options.visible_device_list = "0"
K.set_session(tf.Session(config=config))
# Load model
model_name = "Original"
converter_name = "Masked"
if use_gan:
model_name = "GAN"
converter_name = "GAN"
# -----------------------------------------------------------
# FIXING THE BUG with Model loading:
# model = PluginLoader.get_model(model_name)(Path(self._model_path(use_gan)))
# TypeError: __init__() takes exactly 3 arguments (2 given)
# -----------------------------------------------------------
# tmp_1 = PluginLoader.get_model(model_name)
# tmp_1 = PluginLoader._import("Model_LIVE", "Model_LIVE") # that works (crutch however)
# tmp_2 = Path(self._model_path(use_gan)) # models/emma_to_jade
# print('\n\n\n{}\n{}\n{}\n{}\n\n\n'.format(tmp_1, type(tmp_1), tmp_2, type(tmp_2)))
# sys.exit(0)
# values in faceit_live module:
# plugins.Model_Original.Model
# <type 'classobj'>
# models/emma_to_jade
# <class 'pathlib.PosixPath'>
# values here:
# plugins.Model_Original.Model.Model
# <type 'classobj'>
# models/emma_to_jade
# <class 'pathlib.PosixPath'>
# -----------------------------------------------------------
# model = PluginLoader.get_model(model_name)(Path(self._model_path(use_gan))) # ==> crash
model = PluginLoader._import("Model_LIVE", "Model_LIVE")(Path(
self._model_path(use_gan)))
# print('\n\n\n{}\n\n\n'.format(self._model_path(use_gan))) # e.g. models/test_2_faces
# sys.exit(0)
if not model.load(swap_model):
print(
'model Not Found! A valid model must be provided to continue!')
exit(1)
print('Checkpoint_1 ... Model loaded')
# -----------------------------------------------------------
# FIXING THE BUG with Converter loading:
# -----------------------------------------------------------
# tmp_1 = PluginLoader.get_converter(converter_name)
# tmp_1 = PluginLoader._import("Convert", "Convert_Masked_LIVE")
# print('\n\n\n{}\n{}\n\n\n'.format(tmp_1, type(tmp_1)))
# sys.exit(0)
# faceit_live module:
# plugins.Convert_Masked.Convert
# <type 'classobj'>
# here:
# plugins.Convert_Masked.Convert
# <type 'classobj'>
# -----------------------------------------------------------
# Load converter
# converter = PluginLoader.get_converter(converter_name) # ==> crash
converter = PluginLoader._import("Convert", "Convert_Masked_LIVE")
converter = converter(model.converter(False),
blur_size=8,
seamless_clone=True,
mask_type="facehullandrect",
erosion_kernel_size=None,
smooth_mask=True,
avg_color_adjust=True)
print('Checkpoint_2 ... Converter loaded')
# Load face filter
filter_person = self._person_a
if swap_model:
filter_person = self._person_b
filter = FaceFilter_LIVE(self._people[filter_person]['faces'])
# Define conversion method per frame
def _convert_frame(frame, convert_colors=True):
# if convert_colors:
# frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) # Swap RGB to BGR to work with OpenCV
DEBUG_MODE = 0
for face in detect_faces_LIVE(frame, "cnn"):
if DEBUG_MODE:
print('Got face!')
# print(dir(face)) # image, x, y, w, h, landmarks
print('Face geometry: ({},{},{},{})'.format(
face.x, face.y, face.w, face.h))
print('Face landmarks: {}'.format(face.landmarks))
cv2.imshow('Face', face.image)
continue
if (not face_filter) or (face_filter and filter.check(face)):
# if 1:
# print(dir(face.landmarks))
# face.landmarks = []
frame = converter.patch_image(frame, face)
if not live:
frame = frame.astype(numpy.float32)
# if convert_colors:
# frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) # Swap RGB to BGR to work with OpenCV
return frame
def _convert_helper(get_frame, t):
return _convert_frame(get_frame(t))
# ===================================================
if live:
print('Staring live mode ...')
print('Press "Q" to Quit')
PATH_TO_VIDEO = './data/videos/emma_360_cut.mp4'
if TEST_2_FACES_FLAG:
# PATH_TO_VIDEO = './_data/videos/pair_360p_original.mp4'
PATH_TO_VIDEO = './data/videos/pair_360p_cut.mp4'
video_capture = cv2.VideoCapture(PATH_TO_VIDEO)
width = video_capture.get(3) # float
height = video_capture.get(4) # float
print("video dimensions = {} x {}".format(width, height))
while 1:
ret, frame = video_capture.read()
# print(frame.shape, frame.dtype) # (360, 640, 3), uint8
# frame = cv2.resize(frame, (640, 480))
print('HANDLING NEW FRAME ...')
if CROP_HALF_OF_FRAME == 'left':
frame[:, 0:frame.shape[1] /
2] = 0 # ~ cropping left half of an image
# elif CROP_HALF_OF_FRAME == 'right':
# pass
if not ret:
print("RET IS NONE ... I'M QUIT")
video_capture.release()
break
# block without try/except - to catch actual errors:
frame = cv2.flip(frame, 1)
image = _convert_frame(frame, convert_colors=False)
print('GOT AN IMAGE!')
frame = cv2.flip(frame, 1)
image = cv2.flip(image, 1)
try: # with flip:
# flip image, because webcam inverts it and we trained the model the other way!
frame = cv2.flip(frame, 1)
image = _convert_frame(frame, convert_colors=False)
print('GOT AN IMAGE!')
# flip it back
frame = cv2.flip(frame, 1)
image = cv2.flip(image, 1)
except:
try: # without flip:
image = _convert_frame(frame, convert_colors=False)
print('GOT AN IMAGE!')
except:
print("HMM ... CONVERTATION FAILED ... I'M QUIT")
continue
# video_capture.release()
# break
cv2.imshow('Video', image)
cv2.imshow('Original', frame)
# Hit 'q' on the keyboard to quit!
if cv2.waitKey(1) & 0xFF == ord('q'):
print("KEYBOARD INTERRUPT ... I'M QUIT")
video_capture.release()
break
cv2.destroyAllWindows()
exit()
# ===================================================
media_path = self._video_path({'name': video_file})
if not photos:
# Process video; start loading the video clip
video = VideoFileClip(media_path)
# If a duration is set, trim clip
if duration:
video = video.subclip(start_time, start_time + duration)
# Resize clip before processing
if width:
video = video.resize(width=width)
# Crop clip if desired
if crop_x:
video = video.fx(crop, x2=video.w / 2)
# Kick off convert frames for each frame
new_video = video.fl(_convert_helper)
# Stack clips side by side
if side_by_side:
def add_caption(caption, clip):
text = (TextClip(caption,
font='Amiri-regular',
color='white',
fontsize=80).margin(40).set_duration(
clip.duration).on_color(
color=(0, 0, 0), col_opacity=0.6))
return CompositeVideoClip([clip, text])
video = add_caption("Original", video)
new_video = add_caption("Swapped", new_video)
final_video = clips_array([[video], [new_video]])
else:
final_video = new_video
# Resize clip after processing
#final_video = final_video.resize(width = (480 * 2))
# Write video
if not os.path.exists(os.path.join(self.OUTPUT_PATH)):
os.makedirs(self.OUTPUT_PATH)
output_path = os.path.join(self.OUTPUT_PATH, video_file)
final_video.write_videofile(output_path, rewrite_audio=True)
# Clean up
del video
del new_video
del final_video
else:
# Process a directory of photos
for face_file in os.listdir(media_path):
face_path = os.path.join(media_path, face_file)
image = cv2.imread(face_path)
image = _convert_frame(image, convert_colors=False)
cv2.imwrite(os.path.join(self.OUTPUT_PATH, face_file), image)
def convert(self,
video_file,
swap_model=False,
duration=None,
start_time=None,
use_gan=False,
face_filter=False,
photos=True,
crop_x=None,
width=None,
side_by_side=False):
# Magic incantation to not have tensorflow blow up with an out of memory error.
import tensorflow as tf
import keras.backend.tensorflow_backend as K
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
config.gpu_options.visible_device_list = "0"
K.set_session(tf.Session(config=config))
# Load model
model_name = "Original"
converter_name = "Masked"
if use_gan:
model_name = "GAN"
converter_name = "GAN"
model = PluginLoader.get_model(model_name)(Path(
self._model_path(use_gan)))
if not model.load(swap_model):
print(
'model Not Found! A valid model must be provided to continue!')
exit(1)
# Load converter
converter = PluginLoader.get_converter(converter_name)
converter = converter(model.converter(False),
blur_size=8,
seamless_clone=True,
mask_type="facehullandrect",
erosion_kernel_size=None,
smooth_mask=True,
avg_color_adjust=True)
# Load face filter
filter_person = self._person_a
if swap_model:
filter_person = self._person_b
filter = FaceFilter(self._people[filter_person]['faces'])
# Define conversion method per frame
def _convert_frame(frame, convert_colors=True):
if convert_colors:
frame = cv2.cvtColor(
frame,
cv2.COLOR_BGR2RGB) # Swap RGB to BGR to work with OpenCV
for face in detect_faces(frame, "cnn"):
if (not face_filter) or (face_filter and filter.check(face)):
frame = converter.patch_image(frame, face)
frame = frame.astype(numpy.float32)
if convert_colors:
frame = cv2.cvtColor(
frame,
cv2.COLOR_BGR2RGB) # Swap RGB to BGR to work with OpenCV
return frame
def _convert_helper(get_frame, t):
return _convert_frame(get_frame(t))
media_path = self._video_path({'name': video_file})
if not photos:
# Process video; start loading the video clip
video = VideoFileClip(media_path)
# If a duration is set, trim clip
if duration:
video = video.subclip(start_time, start_time + duration)
# Resize clip before processing
if width:
video = video.resize(width=width)
# Crop clip if desired
if crop_x:
video = video.fx(crop, x2=video.w / 2)
# Kick off convert frames for each frame
new_video = video.fl(_convert_helper)
# Stack clips side by side
if side_by_side:
def add_caption(caption, clip):
text = (TextClip(caption,
font='Amiri-regular',
color='white',
fontsize=80).margin(40).set_duration(
clip.duration).on_color(
color=(0, 0, 0), col_opacity=0.6))
return CompositeVideoClip([clip, text])
video = add_caption("Original", video)
new_video = add_caption("Swapped", new_video)
final_video = clips_array([[video], [new_video]])
else:
final_video = new_video
# Resize clip after processing
#final_video = final_video.resize(width = (480 * 2))
# Write video
output_path = os.path.join(self.OUTPUT_PATH, video_file)
final_video.write_videofile(output_path, rewrite_audio=True)
# Clean up
del video
del new_video
del final_video
else:
# Process a directory of photos
for face_file in os.listdir(media_path):
face_path = os.path.join(media_path, face_file)
image = cv2.imread(face_path)
image = _convert_frame(image, convert_colors=False)
cv2.imwrite(os.path.join(self.OUTPUT_PATH, face_file), image)
class MovieStim3(BaseVisualStim, ContainerMixin, TextureMixin):
"""A stimulus class for playing movies.
This class uses MoviePy and FFMPEG as a backend for loading and decoding
video data from files.
Parameters
----------
filename : str
A string giving the relative or absolute path to the movie.
flipVert : True or *False*
If True then the movie will be top-bottom flipped
flipHoriz : True or *False*
If True then the movie will be right-left flipped
volume :
The nominal level is 100, and 0 is silence.
loop : bool, optional
Whether to start the movie over from the beginning if draw is called and
the movie is done.
Examples
--------
See Movie2Stim.py for demo.
"""
def __init__(self,
win,
filename="",
units='pix',
size=None,
pos=(0.0, 0.0),
anchor="center",
ori=0.0,
flipVert=False,
flipHoriz=False,
color=(1.0, 1.0, 1.0),
colorSpace='rgb',
opacity=1.0,
volume=1.0,
name='',
loop=False,
autoLog=True,
depth=0.0,
noAudio=False,
vframe_callback=None,
fps=None,
interpolate=True):
# what local vars are defined (these are the init params) for use
# by __repr__
self._initParams = dir()
self._initParams.remove('self')
super(MovieStim3, self).__init__(win,
units=units,
name=name,
autoLog=False)
retraceRate = win._monitorFrameRate
if retraceRate is None:
retraceRate = win.getActualFrameRate()
if retraceRate is None:
logging.warning("FrameRate could not be supplied by psychopy; "
"defaulting to 60.0")
retraceRate = 60.0
self._retraceInterval = 1.0 / retraceRate
self.filename = pathToString(filename)
self.loop = loop
self.flipVert = flipVert
self.flipHoriz = flipHoriz
self.pos = numpy.asarray(pos, float)
self.anchor = anchor
self.depth = depth
self.opacity = opacity
self.interpolate = interpolate
self.noAudio = noAudio
self._audioStream = None
self.useTexSubImage2D = True
if noAudio: # to avoid dependency problems in silent movies
self.sound = None
else:
from psychopy import sound
self.sound = sound
# set autoLog (now that params have been initialised)
self.autoLog = autoLog
if autoLog:
logging.exp("Created %s = %s" % (self.name, str(self)))
self._videoClock = Clock()
self.loadMovie(self.filename)
self.setVolume(volume)
self.nDroppedFrames = 0
# size
if size is None:
self.size = numpy.array([self._mov.w, self._mov.h], float)
else:
self.size = val2array(size)
self.ori = ori
self._updateVertices()
@property
def interpolate(self):
"""Enable linear interpolation (`bool').
If `True` linear filtering will be applied to the video making the image
less pixelated if scaled.
"""
return self._interpolate
@interpolate.setter
def interpolate(self, value):
self._interpolate = value
self._texFilterNeedsUpdate = True
@property
def duration(self):
"""Duration of the video clip in seconds (`float`). Only valid after
loading a clip, always returning `0.0` if not.
"""
if self._mov is None:
return 0.0
return self._mov.duration
@property
def frameInterval(self):
"""Time in seconds each frame is to be presented on screen (`float`).
Value is `0.0` if no movie is loaded.
"""
if self._mov is None:
return 0.0
return 1. / self._mov.fps
def reset(self):
self._numpyFrame = None
self._nextFrameT = 0.0
self._texID = None
self.status = NOT_STARTED
self.nDroppedFrames = 0
def setMovie(self, filename, log=True):
"""See `~MovieStim.loadMovie` (the functions are identical).
This form is provided for syntactic consistency with other visual
stimuli.
Parameters
----------
filename : str
The name of the file, including path if necessary.
log : bool
Log this event.
"""
self.loadMovie(filename, log=log)
def loadMovie(self, filename, log=True):
"""Load a movie from file.
After the file is loaded `MovieStim.duration` is updated with the movie
duration (in seconds).
Parameters
----------
filename : str
The name of the file, including path if necessary.
log : bool
Log this event.
"""
filename = pathToString(filename)
self.reset() # set status and timestamps etc
# Create Video Stream stuff
if os.path.isfile(filename):
self._mov = VideoFileClip(filename, audio=(1 - self.noAudio))
if (not self.noAudio) and (self._mov.audio is not None):
sound = self.sound
try:
self._audioStream = sound.Sound(
self._mov.audio.to_soundarray(),
sampleRate=self._mov.audio.fps)
except:
# JWE added this as a patch for a moviepy oddity where the
# duration is inflated in the saved file causes the
# audioclip to be the wrong length, so round down and it
# should work
jwe_tmp = self._mov.subclip(0, round(self._mov.duration))
self._audioStream = sound.Sound(
jwe_tmp.audio.to_soundarray(),
sampleRate=self._mov.audio.fps)
del (jwe_tmp)
else: # make sure we set to None (in case prev clip had audio)
self._audioStream = None
else:
raise IOError("Movie file '%s' was not found" % filename)
# mov has attributes:
# size, duration, fps
# mov.audio has attributes
# duration, fps (aka sampleRate), to_soundarray()
self._frameInterval = 1.0 / self._mov.fps
# self.duration = self._mov.duration
self.filename = filename
self._updateFrameTexture()
logAttrib(self, log, 'movie', filename)
def play(self, log=True):
"""Continue a paused movie from current position.
"""
status = self.status
if status != PLAYING:
self.status = PLAYING # moved this to get better audio behavior - JK
# Added extra check to prevent audio doubling - JK
if self._audioStream is not None and self._audioStream.status is not PLAYING:
self._audioStream.play()
if status == PAUSED:
if self.getCurrentFrameTime(
) < 0: # Check for valid timestamp, correct if needed -JK
self._audioSeek(0)
else:
self._audioSeek(self.getCurrentFrameTime())
self._videoClock.reset(-self.getCurrentFrameTime())
if log and self.autoLog:
self.win.logOnFlip("Set %s playing" % (self.name),
level=logging.EXP,
obj=self)
self._updateFrameTexture()
def pause(self, log=True):
"""
Pause the current point in the movie (sound will stop, current frame
will not advance). If play() is called again both will restart.
"""
if self.status == PLAYING:
self.status = PAUSED
if self._audioStream:
if prefs.hardware['audioLib'] in ['sounddevice', 'PTB']:
self._audioStream.pause(
) # sounddevice and PTB have a "pause" function -JK
else:
self._audioStream.stop()
if log and self.autoLog:
self.win.logOnFlip("Set %s paused" % (self.name),
level=logging.EXP,
obj=self)
return True
if log and self.autoLog:
self.win.logOnFlip("Failed Set %s paused" % (self.name),
level=logging.EXP,
obj=self)
return False
def stop(self, log=True):
"""Stop the current point in the movie (sound will stop, current frame
will not advance). Once stopped the movie cannot be restarted -
it must be loaded again. Use pause() if you may need to restart
the movie.
"""
if self.status != STOPPED:
self._unload()
self.reset()
self.status = STOPPED # set status to STOPPED after _unload
if log and self.autoLog:
self.win.logOnFlip("Set %s stopped" % (self.name),
level=logging.EXP,
obj=self)
def setVolume(self, volume):
pass # to do
def setFlipHoriz(self, newVal=True, log=True):
"""If set to True then the movie will be flipped horizontally
(left-to-right). Note that this is relative to the original,
not relative to the current state.
"""
self.flipHoriz = newVal
logAttrib(self, log, 'flipHoriz')
self._needVertexUpdate = True
def setFlipVert(self, newVal=True, log=True):
"""If set to True then the movie will be flipped vertically
(top-to-bottom). Note that this is relative to the original,
not relative to the current state.
"""
self.flipVert = newVal
logAttrib(self, log, 'flipVert')
self._needVertexUpdate = True
def getFPS(self):
"""Get the movie frames per second.
Returns
-------
float
Frames per second.
"""
return float(self._mov.fps)
def getCurrentFrameTime(self):
"""Get the time that the movie file specified the current
video frame as having.
"""
return self._nextFrameT - self.frameInterval
def _updateFrameTexture(self):
"""Update texture pixel store to contain the present frame. Decoded
frame image samples are streamed to the texture buffer.
"""
if self._nextFrameT is None or self._nextFrameT < 0:
# movie has no current position (or invalid position -JK),
# need to reset the clock to zero in order to have the
# timing logic work otherwise the video stream would skip
# frames until the time since creating the movie object has passed
self._videoClock.reset()
self._nextFrameT = 0.0
# only advance if next frame (half of next retrace rate)
if self._nextFrameT > self.duration:
self._onEos()
elif self._numpyFrame is not None:
if self._nextFrameT > (self._videoClock.getTime() -
self._retraceInterval / 2.0):
return None
while self._nextFrameT <= (self._videoClock.getTime() -
self._frameInterval * 2):
self.nDroppedFrames += 1
if self.nDroppedFrames <= reportNDroppedFrames:
logging.warning(
"{}: Video catchup needed, advancing self._nextFrameT from"
" {} to {}".format(self._videoClock.getTime(),
self._nextFrameT,
self._nextFrameT + self._frameInterval))
if self.nDroppedFrames == reportNDroppedFrames:
logging.warning(
"Max reportNDroppedFrames reached, will not log any more dropped frames"
)
self._nextFrameT += self._frameInterval
try:
self._numpyFrame = self._mov.get_frame(self._nextFrameT)
except OSError:
if self.autoLog:
logging.warning(
"Frame {} not found, moving one frame and trying again".
format(self._nextFrameT),
obj=self)
self._nextFrameT += self._frameInterval
self._updateFrameTexture()
useSubTex = self.useTexSubImage2D
if self._texID is None:
self._texID = GL.GLuint()
GL.glGenTextures(1, ctypes.byref(self._texID))
useSubTex = False
GL.glActiveTexture(GL.GL_TEXTURE0)
# bind that name to the target
GL.glBindTexture(GL.GL_TEXTURE_2D, self._texID)
# bind the texture in openGL
GL.glEnable(GL.GL_TEXTURE_2D)
# makes the texture map wrap (this is actually default anyway)
GL.glTexParameteri(GL.GL_TEXTURE_2D, GL.GL_TEXTURE_WRAP_S, GL.GL_CLAMP)
GL.glTexParameteri(GL.GL_TEXTURE_2D, GL.GL_TEXTURE_WRAP_T, GL.GL_CLAMP)
# data from PIL/numpy is packed, but default for GL is 4 bytes
GL.glPixelStorei(GL.GL_UNPACK_ALIGNMENT, 1)
# important if using bits++ because GL_LINEAR
# sometimes extrapolates to pixel vals outside range
if self.interpolate:
GL.glTexParameteri(GL.GL_TEXTURE_2D, GL.GL_TEXTURE_MAG_FILTER,
GL.GL_LINEAR)
GL.glTexParameteri(GL.GL_TEXTURE_2D, GL.GL_TEXTURE_MIN_FILTER,
GL.GL_LINEAR)
if useSubTex is False:
GL.glTexImage2D(GL.GL_TEXTURE_2D, 0, GL.GL_RGB8,
self._numpyFrame.shape[1],
self._numpyFrame.shape[0], 0, GL.GL_RGB,
GL.GL_UNSIGNED_BYTE, self._numpyFrame.ctypes)
else:
GL.glTexSubImage2D(GL.GL_TEXTURE_2D, 0, 0, 0,
self._numpyFrame.shape[1],
self._numpyFrame.shape[0], GL.GL_RGB,
GL.GL_UNSIGNED_BYTE,
self._numpyFrame.ctypes)
else:
GL.glTexParameteri(GL.GL_TEXTURE_2D, GL.GL_TEXTURE_MAG_FILTER,
GL.GL_NEAREST)
GL.glTexParameteri(GL.GL_TEXTURE_2D, GL.GL_TEXTURE_MIN_FILTER,
GL.GL_NEAREST)
if useSubTex is False:
GL.glTexImage2D(GL.GL_TEXTURE_2D, 0, GL.GL_RGB8,
self._numpyFrame.shape[1],
self._numpyFrame.shape[0], 0, GL.GL_BGR,
GL.GL_UNSIGNED_BYTE, self._numpyFrame.ctypes)
else:
GL.glTexSubImage2D(GL.GL_TEXTURE_2D, 0, 0, 0,
self._numpyFrame.shape[1],
self._numpyFrame.shape[0], GL.GL_BGR,
GL.GL_UNSIGNED_BYTE,
self._numpyFrame.ctypes)
GL.glTexEnvi(GL.GL_TEXTURE_ENV, GL.GL_TEXTURE_ENV_MODE,
GL.GL_MODULATE) # ?? do we need this - think not!
if self.status == PLAYING:
self._nextFrameT += self._frameInterval
def draw(self, win=None):
"""Draw the current frame to a particular visual.Window (or to the
default win for this object if not specified). The current position in
the movie will be determined automatically.
This method should be called on every frame that the movie is meant to
appear.
Parameters
----------
win : :class:`~psychopy.visual.Window` or None
Window the video is being drawn to. If `None`, the window specified
by property `win` will be used. Default is `None`.
"""
if (self.status == NOT_STARTED
or (self.status == FINISHED and self.loop)):
self.play()
elif self.status == FINISHED and not self.loop:
return
if win is None:
win = self.win
self._selectWindow(win)
self._updateFrameTexture() # will check if it's needed
# scale the drawing frame and get to centre of field
GL.glPushMatrix() # push before drawing, pop after
# push the data for client attributes
GL.glPushClientAttrib(GL.GL_CLIENT_ALL_ATTRIB_BITS)
self.win.setScale('pix')
# move to centre of stimulus and rotate
vertsPix = self.verticesPix
# bind textures
GL.glActiveTexture(GL.GL_TEXTURE1)
GL.glBindTexture(GL.GL_TEXTURE_2D, 0)
GL.glEnable(GL.GL_TEXTURE_2D)
GL.glActiveTexture(GL.GL_TEXTURE0)
GL.glBindTexture(GL.GL_TEXTURE_2D, self._texID)
GL.glEnable(GL.GL_TEXTURE_2D)
# sets opacity (1,1,1 = RGB placeholder)
GL.glColor4f(1, 1, 1, self.opacity)
array = (GL.GLfloat * 32)(
1,
1, # texture coords
vertsPix[0, 0],
vertsPix[0, 1],
0., # vertex
0,
1,
vertsPix[1, 0],
vertsPix[1, 1],
0.,
0,
0,
vertsPix[2, 0],
vertsPix[2, 1],
0.,
1,
0,
vertsPix[3, 0],
vertsPix[3, 1],
0.,
)
# 2D texture array, 3D vertex array
GL.glInterleavedArrays(GL.GL_T2F_V3F, 0, array)
GL.glDrawArrays(GL.GL_QUADS, 0, 4)
GL.glPopClientAttrib()
GL.glPopMatrix()
# unbind the textures
GL.glActiveTexture(GL.GL_TEXTURE0)
GL.glBindTexture(GL.GL_TEXTURE_2D, 0)
GL.glEnable(GL.GL_TEXTURE_2D) # implicitly disables 1D
def seek(self, t):
"""Go to a specific point in time for both the audio and video streams
"""
# video is easy: set both times to zero and update the frame texture
self._nextFrameT = t
self._videoClock.reset(t)
self._audioSeek(t)
def _audioSeek(self, t):
sound = self.sound
if self._audioStream is None:
return # do nothing
# check if sounddevice or PTB is being used. If so we can use seek. If not we
# have to reload the audio stream and begin at the new loc
if prefs.hardware['audioLib'] in ['sounddevice', 'PTB']:
self._audioStream.seek(t)
else:
self._audioStream.stop()
sndArray = self._mov.audio.to_soundarray()
startIndex = int(t * self._mov.audio.fps)
self._audioStream = sound.Sound(sndArray[startIndex:, :],
sampleRate=self._mov.audio.fps)
if self.status != PAUSED: # Allows for seeking while paused - JK
self._audioStream.play()
def _getAudioStreamTime(self):
return self._audio_stream_clock.getTime()
def _unload(self):
# remove textures from graphics card to prevent crash
self.clearTextures()
if self._mov is not None:
self._mov.close()
self._mov = None
self._numpyFrame = None
if self._audioStream is not None:
self._audioStream.stop()
self._audioStream = None
self.status = FINISHED
def _onEos(self):
if self.loop:
self.seek(0.0)
else:
self.status = FINISHED
self.stop()
if self.autoLog:
self.win.logOnFlip("Set %s finished" % self.name,
level=logging.EXP,
obj=self)
def __del__(self):
try:
self._unload()
except (ImportError, ModuleNotFoundError, TypeError):
pass # has probably been garbage-collected already
def setAutoDraw(self, val, log=None):
"""Add or remove a stimulus from the list of stimuli that will be
automatically drawn on each flip.
Parameters
----------
val : bool
True to add the stimulus to the draw list, False to remove it.
"""
if val:
self.play(log=False) # set to play in case stopped
else:
self.pause(log=False)
# add to drawing list and update status
setAttribute(self, 'autoDraw', val, log)
if file_.endswith('mp4'):
files_to_upload = []
# 3) Get video size
video_path = join(conf.VIDEO_DIR, file_)
size_in_bytes = getsize(video_path)
if size_in_bytes > conf.MAX_TRANSACTION_SIZE:
# 4) If the size exceeds the transaction limit split it into chunks
num_splits = ceil(size_in_bytes /
conf.MAX_TRANSACTION_SIZE)
video = VideoFileClip(video_path)
for i, (t_start, t_end) in enumerate(
get_next_interval(video.duration, num_splits)):
video_clip_path = join(
conf.VIDEO_DIR, f'{file_[0:-4]}_part{i}.mp4')
video_clip = video.subclip(t_start, t_end)
video_clip.write_videofile(video_clip_path,
codec='libx265')
files_to_upload.append(video_clip_path)
else:
files_to_upload.append(video_path)
# 4) Upload the video and after doing so move it to the uploaded folder
for file_to_upload in files_to_upload:
print(f'Uploading {path.basename(file_to_upload)}...')
# upload video to IPFS
res = ipfs_client.add(file_to_upload)
# outputs IPFS hash (=hash of file content)
def convert(self,
video_file,
swap_model=False,
duration=None,
start_time=None,
use_gan=False,
face_filter=False,
photos=True,
crop_x=None,
width=None,
side_by_side=False,
live=False,
webcam=False):
# Magic incantation to not have tensorflow blow up with an out of memory error.
import tensorflow as tf
import keras.backend.tensorflow_backend as K
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
config.gpu_options.visible_device_list = "0"
K.set_session(tf.Session(config=config))
# Load model
model_name = "Original"
converter_name = "Masked"
if use_gan:
model_name = "GAN"
converter_name = "GAN"
model = PluginLoader.get_model(model_name)(Path(
self._model_path(use_gan)))
if not model.load(swap_model):
print(
'model Not Found! A valid model must be provided to continue!')
exit(1)
# Load converter
converter = PluginLoader.get_converter(converter_name)
converter = converter(model.converter(False),
blur_size=8,
seamless_clone=True,
mask_type="facehullandrect",
erosion_kernel_size=None,
smooth_mask=True,
avg_color_adjust=True)
# Load face filter
filter_person = self._person_a
if swap_model:
filter_person = self._person_b
filter = FaceFilter(self._people[filter_person]['faces'])
# Define conversion method per frame
def _convert_frame(frame, convert_colors=True):
if convert_colors:
frame = cv2.cvtColor(
frame,
cv2.COLOR_BGR2RGB) # Swap RGB to BGR to work with OpenCV
for face in detect_faces(frame, "cnn"):
if (not face_filter) or (face_filter and filter.check(face)):
frame = converter.patch_image(frame, face)
if (not live and not webcam):
frame = frame.astype(numpy.float32)
if convert_colors:
frame = cv2.cvtColor(
frame,
cv2.COLOR_BGR2RGB) # Swap RGB to BGR to work with OpenCV
return frame
def _convert_helper(get_frame, t):
return _convert_frame(get_frame(t))
if (live):
# generate dummy content for testing /dev/video1
#ffmpeg -f x11grab -s 640x480 -i :0.0+10,20 -vf format=pix_fmts=yuv420p -f v4l2 /dev/video1
print("Staring live mode. Capturing video from webcam!")
print("Press q to Quit")
# connect to webcam
video_capture = cv2.VideoCapture(0)
time.sleep(1)
width = video_capture.get(3) # float
height = video_capture.get(4) # float
print("webcam dimensions = {} x {}".format(width, height))
#video_capture = cv2.VideoCapture('./data/videos/ale.mp4')
if (webcam):
# create fake webcam device
camera = pyfakewebcam.FakeWebcam('/dev/video1', 640, 480)
camera.print_capabilities()
print(
"Fake webcam created, try using appear.in on Firefox or ")
# loop until user clicks 'q' to exit
while True:
ret, frame = video_capture.read()
frame = cv2.resize(frame, (640, 480))
# flip image, because webcam inverts it and we trained the model the other way!
frame = cv2.flip(frame, 1)
image = _convert_frame(frame, convert_colors=False)
# flip it back
image = cv2.flip(image, 1)
if (webcam):
time.sleep(1 / 30.0)
# firefox needs RGB
# image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
# chrome and skype UYUV - not working at the moment
# image = cv2.cvtColor(image, cv2.COLOR_BGR2YUV)
camera.schedule_frame(image)
#print("writing to stream")
else:
cv2.imshow('Video', image)
#print("writing to screen")
# Hit 'q' on the keyboard to quit!
if cv2.waitKey(1) & 0xFF == ord('q'):
video_capture.release()
break
cv2.destroyAllWindows()
exit()
media_path = self._video_path({'name': video_file})
if not photos:
# Process video; start loading the video clip
video = VideoFileClip(media_path)
# If a duration is set, trim clip
if duration:
video = video.subclip(start_time, start_time + duration)
# Resize clip before processing
if width:
video = video.resize(width=width)
# Crop clip if desired
if crop_x:
video = video.fx(crop, x2=video.w / 2)
# Kick off convert frames for each frame
new_video = video.fl(_convert_helper)
# Stack clips side by side
if side_by_side:
def add_caption(caption, clip):
text = (TextClip(caption,
font='Amiri-regular',
color='white',
fontsize=80).margin(40).set_duration(
clip.duration).on_color(
color=(0, 0, 0), col_opacity=0.6))
return CompositeVideoClip([clip, text])
video = add_caption("Original", video)
new_video = add_caption("Swapped", new_video)
final_video = clips_array([[video], [new_video]])
else:
final_video = new_video
# Resize clip after processing
#final_video = final_video.resize(width = (480 * 2))
# Write video
if not os.path.exists(os.path.join(self.OUTPUT_PATH)):
os.makedirs(self.OUTPUT_PATH)
output_path = os.path.join(self.OUTPUT_PATH, video_file)
final_video.write_videofile(output_path, rewrite_audio=True)
# Clean up
del video
del new_video
del final_video
else:
# Process a directory of photos
for face_file in os.listdir(media_path):
face_path = os.path.join(media_path, face_file)
image = cv2.imread(face_path)
image = _convert_frame(image, convert_colors=False)
cv2.imwrite(os.path.join(self.OUTPUT_PATH, face_file), image)
class MovieStim3(BaseVisualStim, ContainerMixin):
"""A stimulus class for playing movies (mpeg, avi, etc...) in PsychoPy
that does not require avbin. Instead it requires the cv2 python package
for OpenCV. The VLC media player also needs to be installed on the
psychopy computer.
**Example**::
See Movie2Stim.py for demo.
"""
def __init__(self, win,
filename="",
units='pix',
size=None,
pos=(0.0, 0.0),
ori=0.0,
flipVert=False,
flipHoriz=False,
color=(1.0, 1.0, 1.0),
colorSpace='rgb',
opacity=1.0,
volume=1.0,
name='',
loop=False,
autoLog=True,
depth=0.0,
noAudio=False,
vframe_callback=None,
fps=None,
interpolate=True):
"""
:Parameters:
filename :
a string giving the relative or absolute path to the movie.
flipVert : True or *False*
If True then the movie will be top-bottom flipped
flipHoriz : True or *False*
If True then the movie will be right-left flipped
volume :
The nominal level is 100, and 0 is silence.
loop : bool, optional
Whether to start the movie over from the beginning if draw is
called and the movie is done.
"""
# what local vars are defined (these are the init params) for use
# by __repr__
self._initParams = dir()
self._initParams.remove('self')
super(MovieStim3, self).__init__(win, units=units, name=name,
autoLog=False)
retraceRate = win._monitorFrameRate
if retraceRate is None:
retraceRate = win.getActualFrameRate()
if retraceRate is None:
logging.warning("FrameRate could not be supplied by psychopy; "
"defaulting to 60.0")
retraceRate = 60.0
self._retraceInterval = old_div(1.0, retraceRate)
self.filename = filename
self.loop = loop
self.flipVert = flipVert
self.flipHoriz = flipHoriz
self.pos = numpy.asarray(pos, float)
self.depth = depth
self.opacity = float(opacity)
self.interpolate = interpolate
self.noAudio = noAudio
self._audioStream = None
self.useTexSubImage2D = True
if noAudio: # to avoid dependency problems in silent movies
self.sound = None
else:
from psychopy import sound
self.sound = sound
self._videoClock = Clock()
self.loadMovie(self.filename)
self.setVolume(volume)
self.nDroppedFrames = 0
# size
if size is None:
self.size = numpy.array([self._mov.w, self._mov.h],
float)
else:
self.size = val2array(size)
self.ori = ori
self._updateVertices()
# set autoLog (now that params have been initialised)
self.autoLog = autoLog
if autoLog:
logging.exp("Created %s = %s" % (self.name, str(self)))
def reset(self):
self._numpyFrame = None
self._nextFrameT = None
self._texID = None
self.status = NOT_STARTED
def setMovie(self, filename, log=True):
"""See `~MovieStim.loadMovie` (the functions are identical).
This form is provided for syntactic consistency with other visual
stimuli.
"""
self.loadMovie(filename, log=log)
def loadMovie(self, filename, log=True):
"""Load a movie from file
:Parameters:
filename: string
The name of the file, including path if necessary
After the file is loaded MovieStim.duration is updated with the movie
duration (in seconds).
"""
self.reset() # set status and timestamps etc
# Create Video Stream stuff
if os.path.isfile(filename):
self._mov = VideoFileClip(filename, audio=(1 - self.noAudio))
if (not self.noAudio) and (self._mov.audio is not None):
sound = self.sound
try:
self._audioStream = sound.Sound(
self._mov.audio.to_soundarray(),
sampleRate=self._mov.audio.fps)
except:
# JWE added this as a patch for a moviepy oddity where the
# duration is inflated in the saved file causes the
# audioclip to be the wrong length, so round down and it
# should work
jwe_tmp = self._mov.subclip(0, round(self._mov.duration))
self._audioStream = sound.Sound(
jwe_tmp.audio.to_soundarray(),
sampleRate=self._mov.audio.fps)
del(jwe_tmp)
else: # make sure we set to None (in case prev clip had audio)
self._audioStream = None
else:
raise IOError("Movie file '%s' was not found" % filename)
# mov has attributes:
# size, duration, fps
# mov.audio has attributes
# duration, fps (aka sampleRate), to_soundarray()
self._frameInterval = old_div(1.0, self._mov.fps)
self.duration = self._mov.duration
self.filename = filename
self._updateFrameTexture()
logAttrib(self, log, 'movie', filename)
def play(self, log=True):
"""Continue a paused movie from current position.
"""
status = self.status
if status != PLAYING:
if self._audioStream is not None:
self._audioStream.play()
if status == PAUSED:
if self.getCurrentFrameTime() < 0:
self._audioSeek(0)
else:
self._audioSeek(self.getCurrentFrameTime())
self.status = PLAYING
self._videoClock.reset(-self.getCurrentFrameTime())
if log and self.autoLog:
self.win.logOnFlip("Set %s playing" % (self.name),
level=logging.EXP, obj=self)
self._updateFrameTexture()
def pause(self, log=True):
"""
Pause the current point in the movie (sound will stop, current frame
will not advance). If play() is called again both will restart.
"""
if self.status == PLAYING:
self.status = PAUSED
if self._audioStream:
self._audioStream.stop()
if log and self.autoLog:
self.win.logOnFlip("Set %s paused" %
(self.name), level=logging.EXP, obj=self)
return True
if log and self.autoLog:
self.win.logOnFlip("Failed Set %s paused" %
(self.name), level=logging.EXP, obj=self)
return False
def stop(self, log=True):
"""Stop the current point in the movie (sound will stop, current frame
will not advance). Once stopped the movie cannot be restarted -
it must be loaded again. Use pause() if you may need to restart
the movie.
"""
if self.status != STOPPED:
self.status = STOPPED
self._unload()
self.reset()
if log and self.autoLog:
self.win.logOnFlip("Set %s stopped" % (self.name),
level=logging.EXP, obj=self)
def setVolume(self, volume):
pass # to do
def setFlipHoriz(self, newVal=True, log=True):
"""If set to True then the movie will be flipped horizontally
(left-to-right). Note that this is relative to the original,
not relative to the current state.
"""
self.flipHoriz = newVal
logAttrib(self, log, 'flipHoriz')
self._needVertexUpdate = True
def setFlipVert(self, newVal=True, log=True):
"""If set to True then the movie will be flipped vertically
(top-to-bottom). Note that this is relative to the original,
not relative to the current state.
"""
self.flipVert = newVal
logAttrib(self, log, 'flipVert')
self._needVertexUpdate = True
def getFPS(self):
"""
Returns the movie frames per second playback speed.
"""
return self._mov.fps
def getCurrentFrameTime(self):
"""Get the time that the movie file specified the current
video frame as having.
"""
return self._nextFrameT - self._frameInterval
def _updateFrameTexture(self):
if self._nextFrameT is None:
# movie has no current position, need to reset the clock
# to zero in order to have the timing logic work
# otherwise the video stream would skip frames until the
# time since creating the movie object has passed
self._videoClock.reset()
self._nextFrameT = 0
# only advance if next frame (half of next retrace rate)
if self._nextFrameT > self.duration:
self._onEos()
elif self._numpyFrame is not None:
if self._nextFrameT > (self._videoClock.getTime() -
old_div(self._retraceInterval, 2.0)):
return None
self._numpyFrame = self._mov.get_frame(self._nextFrameT)
useSubTex = self.useTexSubImage2D
if self._texID is None:
self._texID = GL.GLuint()
GL.glGenTextures(1, ctypes.byref(self._texID))
useSubTex = False
# bind the texture in openGL
GL.glEnable(GL.GL_TEXTURE_2D)
# bind that name to the target
GL.glBindTexture(GL.GL_TEXTURE_2D, self._texID)
# makes the texture map wrap (this is actually default anyway)
GL.glTexParameteri(
GL.GL_TEXTURE_2D, GL.GL_TEXTURE_WRAP_S, GL.GL_REPEAT)
# data from PIL/numpy is packed, but default for GL is 4 bytes
GL.glPixelStorei(GL.GL_UNPACK_ALIGNMENT, 1)
# important if using bits++ because GL_LINEAR
# sometimes extrapolates to pixel vals outside range
if self.interpolate:
GL.glTexParameteri(
GL.GL_TEXTURE_2D, GL.GL_TEXTURE_MAG_FILTER, GL.GL_LINEAR)
GL.glTexParameteri(
GL.GL_TEXTURE_2D, GL.GL_TEXTURE_MIN_FILTER, GL.GL_LINEAR)
if useSubTex is False:
GL.glTexImage2D(GL.GL_TEXTURE_2D, 0, GL.GL_RGB8,
self._numpyFrame.shape[1],
self._numpyFrame.shape[0], 0,
GL.GL_RGB, GL.GL_UNSIGNED_BYTE,
self._numpyFrame.ctypes)
else:
GL.glTexSubImage2D(GL.GL_TEXTURE_2D, 0, 0, 0,
self._numpyFrame.shape[1],
self._numpyFrame.shape[0],
GL.GL_RGB, GL.GL_UNSIGNED_BYTE,
self._numpyFrame.ctypes)
else:
GL.glTexParameteri(
GL.GL_TEXTURE_2D, GL.GL_TEXTURE_MAG_FILTER, GL.GL_NEAREST)
GL.glTexParameteri(
GL.GL_TEXTURE_2D, GL.GL_TEXTURE_MIN_FILTER, GL.GL_NEAREST)
if useSubTex is False:
GL.glTexImage2D(GL.GL_TEXTURE_2D, 0, GL.GL_RGB8,
self._numpyFrame.shape[1],
self._numpyFrame.shape[0], 0,
GL.GL_BGR, GL.GL_UNSIGNED_BYTE,
self._numpyFrame.ctypes)
else:
GL.glTexSubImage2D(GL.GL_TEXTURE_2D, 0, 0, 0,
self._numpyFrame.shape[1],
self._numpyFrame.shape[0],
GL.GL_BGR, GL.GL_UNSIGNED_BYTE,
self._numpyFrame.ctypes)
GL.glTexEnvi(GL.GL_TEXTURE_ENV, GL.GL_TEXTURE_ENV_MODE,
GL.GL_MODULATE) # ?? do we need this - think not!
if not self.status == PAUSED:
self._nextFrameT += self._frameInterval
def draw(self, win=None):
"""Draw the current frame to a particular visual.Window (or to the
default win for this object if not specified). The current
position in the movie will be determined automatically.
This method should be called on every frame that the movie is
meant to appear.
"""
if (self.status == NOT_STARTED or
(self.status == FINISHED and self.loop)):
self.play()
elif self.status == FINISHED and not self.loop:
return
if win is None:
win = self.win
self._selectWindow(win)
self._updateFrameTexture() # will check if it's needed
# scale the drawing frame and get to centre of field
GL.glPushMatrix() # push before drawing, pop after
# push the data for client attributes
GL.glPushClientAttrib(GL.GL_CLIENT_ALL_ATTRIB_BITS)
self.win.setScale('pix')
# move to centre of stimulus and rotate
vertsPix = self.verticesPix
# bind textures
GL.glActiveTexture(GL.GL_TEXTURE1)
GL.glBindTexture(GL.GL_TEXTURE_2D, 0)
GL.glEnable(GL.GL_TEXTURE_2D)
GL.glActiveTexture(GL.GL_TEXTURE0)
GL.glBindTexture(GL.GL_TEXTURE_2D, self._texID)
GL.glEnable(GL.GL_TEXTURE_2D)
# sets opacity (1,1,1 = RGB placeholder)
GL.glColor4f(1, 1, 1, self.opacity)
array = (GL.GLfloat * 32)(
1, 1, # texture coords
vertsPix[0, 0], vertsPix[0, 1], 0., # vertex
0, 1,
vertsPix[1, 0], vertsPix[1, 1], 0.,
0, 0,
vertsPix[2, 0], vertsPix[2, 1], 0.,
1, 0,
vertsPix[3, 0], vertsPix[3, 1], 0.,
)
# 2D texture array, 3D vertex array
GL.glInterleavedArrays(GL.GL_T2F_V3F, 0, array)
GL.glDrawArrays(GL.GL_QUADS, 0, 4)
GL.glPopClientAttrib()
GL.glPopAttrib()
GL.glPopMatrix()
# unbind the textures
GL.glActiveTexture(GL.GL_TEXTURE0)
GL.glBindTexture(GL.GL_TEXTURE_2D, 0)
GL.glEnable(GL.GL_TEXTURE_2D) # implicitly disables 1D
def seek(self, t):
"""Go to a specific point in time for both the audio and video streams
"""
# video is easy: set both times to zero and update the frame texture
self._nextFrameT = t
self._videoClock.reset(t)
self._audioSeek(t)
def _audioSeek(self, t):
sound = self.sound
# for sound we need to extract the array again and just begin at new
# loc
if self._audioStream is None:
return # do nothing
self._audioStream.stop()
sndArray = self._mov.audio.to_soundarray()
startIndex = int(t * self._mov.audio.fps)
self._audioStream = sound.Sound(
sndArray[startIndex:, :], sampleRate=self._mov.audio.fps)
self._audioStream.play()
def _getAudioStreamTime(self):
return self._audio_stream_clock.getTime()
def _unload(self):
try:
# remove textures from graphics card to prevent crash
self.clearTextures()
except Exception:
pass
self._mov = None
self._numpyFrame = None
self._audioStream = None
self.status = FINISHED
def _onEos(self):
if self.loop:
self.seek(0.0)
else:
self.status = FINISHED
self.stop()
if self.autoLog:
self.win.logOnFlip("Set %s finished" % self.name,
level=logging.EXP, obj=self)
def __del__(self):
self._unload()
def setAutoDraw(self, val, log=None):
"""Add or remove a stimulus from the list of stimuli that will be
automatically drawn on each flip
:parameters:
- val: True/False
True to add the stimulus to the draw list, False to remove it
"""
if val:
self.play(log=False) # set to play in case stopped
else:
self.pause(log=False)
# add to drawing list and update status
setAttribute(self, 'autoDraw', val, log)
def add_word(self,
word,
collection,
start,
end,
name,
add_type,
word_type,
group,
word_id='',
wordset_id=''):
clean_word = word.strip()
puresave_filename = name.split('.')[0] + "~" + clean_word
# row = {
# "videaname":puresave_filename,
# "wordbase_type":"video"
# }
# w = WordbaseHelper()
# w.init_word(row,clean_word)
# w.insert(row,collection)
data = {
'filename': puresave_filename,
'wordbase_collection': collection,
'word': word,
'add_type': add_type,
'word_type': word_type,
'group': group,
'word_id': word_id,
'wordset_id': wordset_id,
}
work_dir = "D:\BaiduYunDownload"
file_path = ""
parent_path = ""
double_loop_flag = False
for parent, dirnames, filenames in os.walk(work_dir, followlinks=True):
for filename in filenames:
if filename == name:
parent_path = parent
file_path = os.path.join(parent, filename)
double_loop_flag = True
break
if double_loop_flag:
break
start_time = float(start)
end_time = float(end)
pure_filename = name.split('.')[0]
subfile_path = os.path.join(parent_path, pure_filename + ".srt")
video_clip = VideoFileClip(file_path)
clip = video_clip.subclip(start_time, end_time)
target = "D:\BaiduYunDownload\\videos\\" + puresave_filename + ".mp4"
clip.write_videofile(target,
codec='libx264',
verbose=False,
audio=True)
video_clip.close()
subtitle = SSAFile.load(subfile_path)
text = '''
1
00:00:00,000 --> 00:00:00,000
'''
temp = SSAFile().from_string(text)
for sub in subtitle:
if sub.start >= start_time * 1000 and sub.end <= end_time * 1000:
text = sub.text.replace(
clean_word, '<c.video-heightlight>' + clean_word + '</c>')
sub.text = text
sub.shift(s=-start_time)
temp.append(sub)
sub_target = "D:\BaiduYunDownload\\videos\\" + puresave_filename
temp.save(sub_target + '.srt')
vtt = WebVTT().from_srt(sub_target + '.srt')
vtt.save(sub_target + '.vtt')
files = {
"video": open(target, "rb"),
"subtitle": open(sub_target + '.vtt', "rb")
}
# print(files)
# r = requests.post('http://127.0.0.1:5000/video', data=data,files=files)
r = requests.post('http://' + server_ip + '/video',
data=data,
files=files)
# print(r.request)
return "true"
def transform_video(self,
input_path,
output_path,
batch_size=4,
start=0,
end=0):
'''
Transform a video to animation version
https://github.com/lengstrom/fast-style-transfer/blob/master/evaluate.py#L21
'''
# Force to None
end = end or None
if not os.path.isfile(input_path):
raise FileNotFoundError(f'{input_path} does not exist')
output_dir = "/".join(output_path.split("/")[:-1])
os.makedirs(output_dir, exist_ok=True)
is_gg_drive = '/drive/' in output_path
temp_file = ''
if is_gg_drive:
# Writing directly into google drive can be inefficient
temp_file = f'tmp_anime.{output_path.split(".")[-1]}'
def transform_and_write(frames, count, writer):
anime_images = denormalize_input(self.transform(frames),
dtype=np.uint8)
for i in range(0, count):
img = np.clip(anime_images[i], 0, 255)
writer.write_frame(img)
video_clip = VideoFileClip(input_path, audio=False)
if start or end:
video_clip = video_clip.subclip(start, end)
video_writer = ffmpeg_writer.FFMPEG_VideoWriter(temp_file
or output_path,
video_clip.size,
video_clip.fps,
codec="libx264",
preset="medium",
bitrate="2000k",
audiofile=input_path,
threads=None,
ffmpeg_params=None)
total_frames = round(video_clip.fps * video_clip.duration)
print(
f'Transfroming video {input_path}, {total_frames} frames, size: {video_clip.size}'
)
batch_shape = (batch_size, video_clip.size[1], video_clip.size[0], 3)
frame_count = 0
frames = np.zeros(batch_shape, dtype=np.float32)
for frame in tqdm(video_clip.iter_frames()):
try:
frames[frame_count] = frame
frame_count += 1
if frame_count == batch_size:
transform_and_write(frames, frame_count, video_writer)
frame_count = 0
except Exception as e:
print(e)
break
# The last frames
if frame_count != 0:
transform_and_write(frames, frame_count, video_writer)
if temp_file:
# move to output path
shutil.move(temp_file, output_path)
print(f'Animation video saved to {output_path}')
video_writer.close()
