Existing tensor decoder has been bifurcated into two seperate gst
element as:
`yoloodv5tensordecoder`: decodes tensors output(masks) from detection-only
models e.g yolov8s.onnx
`yolsegv8tensordecoder`: decoder tensors output(masks and logits) from
segementation models e.g FastSAM or yolov8s-seg
YOLOv8 model have same tensor output format as FastSAM, so for better
generalization rename fastsamtensordecoder to yolotensordecoder. This
also requires code adaptation to support Yolo based model.
The TsMuxStream internal_stream_type field stores the original
'full' stream type (such as Opus), while the stream_type field
stores the value that will actually be written into the MPEG-TS
packets according to the codec mappings. When checking if
input caps are changing stream type, check the original type.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/9430>
This is needed if the LCEVC enhancement data is part of the video stream as SEI
and the demuxer outputs 'lcevc=false' src caps because LCEVC enhancement data is
not stored as a separate stream in the container.
To clarify, 'lcevc=true' just means that the video buffers have LCEVC metadata
attached. Therefore, it is valid to have a stream with LCEVC enhancement data as
SEI with 'lcevc=false' as long as it is not attached as metadata.
This will be needed once we add support for the demuxer to attach LCEVC metadata
to video buffers if it is stored in a separate track.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/9427>
- Add a videoconvert element before the videosink so that the output
works no matter what format gets negotiated (A444_16LE for me)
- Specify a reasonable video format and size with a capsfilter, so
we don't default to something silly like A444_16LE @ 240p.
- Add a timeoverlay element, so it's obvious when stoppping/restarting
the pipeline that the input stream is just picked up again from the
moment the consumer pipeline is restarted.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/9422>
Modify the API to retrieve the tensor meta to check for the dimensions
as well.
Also fix an API mistake, the buffer whose dimensions should be checheck
is the one inside the GstTensor, not another buffer some outside.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/9411>
Similar to h264parse and h265parse, this patch enhances the element to parse
LCEVC enhancement data from SEI, and attach it to output buffers as GstLcevcMeta.
The 'lcevc' field in the output caps is also set to TRUE or FALSE depending on
whether LCEVC data is present or not.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/9365>
Similar to h264parse and h265parse, this patch improves the element to parse
the SEI registered user data from NAL units. The core structure of H266 SEI for
ITU-T T.35 is the same as the other parsers, so we can re-use the same logic.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/9365>
This causes servers such as AWS MediaLive to drop the connection, and
the message flow chart as documented in the "spec" always has the
server sending it first, and the client replying to it on reception of
the Set Peer Bandwidth, which we do since 286a3829b637.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/9360>
Similar to h264parse, this patch enhances the element to parse LCEVC enhancement
data from SEI, and attach it to output buffers as GstLcevcMeta. The 'lcevc'
field in the output caps is also set to TRUE or FALSE depending on whether LCEVC
data is present or not.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/9335>
This reverts commit fd1428d3ebf937d7656c3788df3ef3ff85fecc31.
As reported in #4524, this changes cause regressions. The problem is due to a
bug in how vp9parse interact with parsebin, presenting downstream negotiation of
alignment to work. This revert to being stuck using frame alignment always,
which fortunately works with libvpx, though less efficient.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/9329>
Only the last frame of a super frame should be displayed by default. When
converting from super frame to frame, mark all frames as decode only except the
last one. This fixes vp90-2-22-svc_1280x720_3.ivf conformance test with
stateless decoders such as VA.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8944>
This is for cases where:
* We *do* want to refer to the PCR stream to figure out global positioning, gap
detection, wrapover correction.
* But we do not want to apply any skew correction to the output
This is useful for cases where:
* the input stream has already been clock-corrected (for example with
mpegtslivesrc)
* or where the output doesn't require synchronization against a clock (ex: for
storage)
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/9301>
Change so that the handling of NAL unit that can't be parsed when using
AU alignment is the same as when using NAL alignment, ie drop the data
if it can't be parsed.
If the AU contains more than one NAL unit any correctly parsed NAL unit
in the AU is kept.
Fixes#4436
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8976>
