Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Close av contaner after use #6501

Merged
merged 3 commits into from
Jul 19, 2023
Merged

Close av contaner after use #6501

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
6d66ca5
close av contaner after use
azhavoro Jul 18, 2023
3866091
apply comments
azhavoro Jul 18, 2023
9e6121a
Merge branch 'develop' into az/av_memory_leak
azhavoro Jul 18, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
(<https://github.com/opencv/cvat/pull/6454>)
- \[SDK\] SDK should not change input data in models (<https://github.com/opencv/cvat/pull/6455>)
- 3D job can not be opened in validation mode (<https://github.com/opencv/cvat/pull/6507>)
- Memory leak related to unclosed av container (<https://github.com/opencv/cvat/pull/6501>)

### Security
- TDB
Expand Down
126 changes: 61 additions & 65 deletions cvat/apps/engine/media_extractors.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -429,32 +429,27 @@ def _has_frame(self, i):

return False

def _decode(self, container):
frame_num = 0
for packet in container.demux():
if packet.stream.type == 'video':
def __iter__(self):
with self._get_av_container() as container:
SpecLad marked this conversation as resolved.
Show resolved Hide resolved
stream = container.streams.video[0]
stream.thread_type = 'AUTO'
frame_num = 0
for packet in container.demux(stream):
for image in packet.decode():
frame_num += 1
if self._has_frame(frame_num - 1):
if packet.stream.metadata.get('rotate'):
old_image = image
pts = image.pts
image = av.VideoFrame().from_ndarray(
rotate_image(
image.to_ndarray(format='bgr24'),
360 - int(container.streams.video[0].metadata.get('rotate'))
360 - int(stream.metadata.get('rotate'))
),
format ='bgr24'
)
image.pts = old_image.pts
image.pts = pts
yield (image, self._source_path[0], image.pts)

def __iter__(self):
container = self._get_av_container()
source_video_stream = container.streams.video[0]
source_video_stream.thread_type = 'AUTO'

return self._decode(container)

def get_progress(self, pos):
duration = self._get_duration()
return pos / duration if duration else None
Expand All @@ -465,38 +460,38 @@ def _get_av_container(self):
return av.open(self._source_path[0])

def _get_duration(self):
container = self._get_av_container()
stream = container.streams.video[0]
duration = None
if stream.duration:
duration = stream.duration
else:
# may have a DURATION in format like "01:16:45.935000000"
duration_str = stream.metadata.get("DURATION", None)
tb_denominator = stream.time_base.denominator
if duration_str and tb_denominator:
_hour, _min, _sec = duration_str.split(':')
duration_sec = 60*60*float(_hour) + 60*float(_min) + float(_sec)
duration = duration_sec * tb_denominator
return duration
with self._get_av_container() as container:
stream = container.streams.video[0]
duration = None
if stream.duration:
duration = stream.duration
else:
# may have a DURATION in format like "01:16:45.935000000"
duration_str = stream.metadata.get("DURATION", None)
tb_denominator = stream.time_base.denominator
if duration_str and tb_denominator:
_hour, _min, _sec = duration_str.split(':')
duration_sec = 60*60*float(_hour) + 60*float(_min) + float(_sec)
duration = duration_sec * tb_denominator
return duration

def get_preview(self, frame):
container = self._get_av_container()
stream = container.streams.video[0]
tb_denominator = stream.time_base.denominator
needed_time = int((frame / stream.guessed_rate) * tb_denominator)
container.seek(offset=needed_time, stream=stream)
for packet in container.demux(stream):
for frame in packet.decode():
return self._get_preview(frame.to_image() if not stream.metadata.get('rotate') \
else av.VideoFrame().from_ndarray(
rotate_image(
frame.to_ndarray(format='bgr24'),
360 - int(container.streams.video[0].metadata.get('rotate'))
),
format ='bgr24'
).to_image()
)
with self._get_av_container() as container:
stream = container.streams.video[0]
tb_denominator = stream.time_base.denominator
needed_time = int((frame / stream.guessed_rate) * tb_denominator)
container.seek(offset=needed_time, stream=stream)
for packet in container.demux(stream):
for frame in packet.decode():
return self._get_preview(frame.to_image() if not stream.metadata.get('rotate') \
else av.VideoFrame().from_ndarray(
rotate_image(
frame.to_ndarray(format='bgr24'),
360 - int(container.streams.video[0].metadata.get('rotate'))
),
format ='bgr24'
).to_image()
)

def get_image_size(self, i):
image = (next(iter(self)))[0]
Expand Down Expand Up @@ -700,6 +695,8 @@ def save_as_chunk(
return image_sizes

class Mpeg4ChunkWriter(IChunkWriter):
FORMAT = 'mp4'

def __init__(self, quality=67):
# translate inversed range [1:100] to [0:51]
quality = round(51 * (100 - quality) / 99)
Expand All @@ -722,21 +719,20 @@ def __init__(self, quality=67):
"preset": "ultrafast",
}

def _create_av_container(self, path, w, h, rate, options, f='mp4'):
def _add_video_stream(self, container, w, h, rate, options):
# x264 requires width and height must be divisible by 2 for yuv420p
if h % 2:
h += 1
if w % 2:
w += 1

container = av.open(path, 'w',format=f)
video_stream = container.add_stream(self._codec_name, rate=rate)
video_stream.pix_fmt = "yuv420p"
video_stream.width = w
video_stream.height = h
video_stream.options = options

return container, video_stream
return video_stream

def save_as_chunk(self, images, chunk_path):
if not images:
Expand All @@ -745,16 +741,16 @@ def save_as_chunk(self, images, chunk_path):
input_w = images[0][0].width
input_h = images[0][0].height

output_container, output_v_stream = self._create_av_container(
path=chunk_path,
w=input_w,
h=input_h,
rate=self._output_fps,
options=self._codec_opts,
)
with av.open(chunk_path, 'w', format=self.FORMAT) as output_container:
output_v_stream = self._add_video_stream(
container=output_container,
w=input_w,
h=input_h,
rate=self._output_fps,
options=self._codec_opts,
)

self._encode_images(images, output_container, output_v_stream)
output_container.close()
self._encode_images(images, output_container, output_v_stream)
return [(input_w, input_h)]

@staticmethod
Expand Down Expand Up @@ -797,16 +793,16 @@ def save_as_chunk(self, images, chunk_path):
output_h = input_h // downscale_factor
output_w = input_w // downscale_factor

output_container, output_v_stream = self._create_av_container(
path=chunk_path,
w=output_w,
h=output_h,
rate=self._output_fps,
options=self._codec_opts,
)
with av.open(chunk_path, 'w', format=self.FORMAT) as output_container:
output_v_stream = self._add_video_stream(
container=output_container,
w=output_w,
h=output_h,
rate=self._output_fps,
options=self._codec_opts,
)

self._encode_images(images, output_container, output_v_stream)
output_container.close()
self._encode_images(images, output_container, output_v_stream)
return [(input_w, input_h)]

def _is_archive(path):
Expand Down