Nvurisrcbin to splitmuxsink without decoding/encoding

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Hi friends,

On both platforms Jetson/dGPU I have the following two pipelines running smoothly:

  1. decoding/encoding pipeline:
gst-launch-1.0 nvurisrcbin uri=rtsp://192.168.0.149 rtsp-reconnect-interval=15 ! nvv4l2h264enc ! h264parse ! splitmuxsink location=video%02d.mp4 max-size-time=15000000000
  1. no decoding/no encoding pipeline:
gst-launch-1.0 rtspsrc location=rtsp://192.168.0.149  ! capsfilter caps=application/x-rtp,media=video ! rtph264depay  ! h264parse  ! splitmuxsink location=video_%02d.mp4 max-size-time=150000000000

My question here is twofold:
a. Is there a way to avoid doing decoding/encoding in the first pipeline?
b. How can the second pipeline reconnect to camera after EOS?

Any help is much appreciated.
BR

3

This is related to your source, if the stream in your source is an H264 stream, you don’t have to use the decoding/encoding part.

You need to implement the reconnection code yourself. You can refer to our source code: watch_source_status in sources\apps\apps-common\src\deepstream_source_bin.c.

4

Thanks for your answer.
My source is h264 indeed, how can I fine tune the deepstream bin/elements to avoid decoding/encoding using elements properties or python?

5

You can just use the piepline 2 you attached. What do you mean how can I fine tune the deepstream bin/elements to avoid decoding/encoding using elements properties or python?

6

I know I can use pipeline2, but it doesn’t have the ability to do rtsp-reconnect-interval as nvurisrcbin, nor nvurisrcbin is interchangable with rtspsrc.
In short: I need the functionality of pipeline2 with the ability to reconnect with cameras.
You mentioned watch_source_status source code, but I need it in python. :-/
Thanks a lot

7

OK, I got what you mean. You can refer to our smart-record. It will slice and save the video for you.

8

OK, I know about the smart-record, but would like to know if it can be triggered without any external message. I want to keep the smart record running 7x24 but interrupt it whenever free disk storage reaches a certain limit.
Thanks

9

You can check for disk storage yourself in your app and proactively stop it when it is full.

10

Can recording be triggered using curl instead of kafka?

11

Yes. You can refer to our deepstream-server demo code. All the controls inside are using curl.

12

Cool !
Example curl command to start/stop recordings?
Documentation about HTTP API syntax?

13

You can achieve your needs through the following methods: add and remove the source.
Please refer to our README:opt\nvidia\deepstream\deepstream\sources\apps\sample_apps\deepstream-server\README

14

Thanks for your response.
It’s kind of a workaround what you’re proposing for a “recording only” pipeline.
In case my pipeline is nvmultiurisrcbin → nvinferserver → nvmsgbrokersinkbin so there’s no way to do add/remove source.

15

At present, the nvmultiurisrcbin plugin has not implemented similar functions. If you want to implement it on your end, you need to refer to opt\nvidia\deepstream\deepstream\sources\apps\sample_apps\deepstream-testsr yourself.

smart_record_event_generator
1 Like
16

Thank @yuweiw for your support.

17

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.