Is it possible to separate pre-process/inference/post-process to separate plugin? So it can processed in different thread and the pre/post-process can be accelerated by HW. Such as nnstreamer and deepstream.
Below MR is only for ONNX, is it possible add one adapter layer for more inference engine?
As GStreamer is plugin based, and since there is very little sharable between framework, each framework support should live in their own plugin in my opinion. This maintain the adaptation layer thin and efficient.
Some framework, like Nvidia and some other maintain their own plugin in their own repository. And this situation is perfectly suitable, as without maintenance, these plugin would have no value.
Regards,
Bing
From: gstreamer-devel <[hidden email]>
On Behalf Of Nicolas Dufresne
Sent: 2021年3月13日
1:03
To: Discussion of the development of and with GStreamer <[hidden email]>
Cc: Aaron Boxer <[hidden email]>
Subject: [EXT] Re: deep learning video analytics
Caution: EXT Email
Le vendredi 12 mars 2021 à 08:03 +0000, Bing Song a écrit :
Hi,
Is there any deep leaning video analytics plugins in upstream repo? There are many such kind of project, such as gstinference/nnstreamer/deepstream/gst-video-analytics. Do you think which one is more likely be accepted
by community?
On my todo to review, but I think this is the most promising one today:
Regards,
Bing
_______________________________________________
gstreamer-devel mailing list
gstreamer-devel mailing list
[hidden email]
https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel
_______________________________________________
gstreamer-devel mailing list
[hidden email]
https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel
Locked
