Start and stop of the process

Robotics & Edge Computing Jetson & Embedded Systems Jetson Nano
,
1

Hi,

I am planning to use Jetson nano to record the video as well as edit the video based on input requests from the cloud. Sometimes, I may be required to stop and start the video in a new file. In this case, do you think stopping nano( without shutdown) and starting again is an option or is there any other better option to handle this scenario?

I am asking the question as every time terminating machine(i.e power off) may have impact on OS drivers or may not close video recording properly.

Thanks
Karun

3

Hi,
If your application is based on gstreamer, you may check if splitmuxsink fits your usecase:
https://gstreamer.freedesktop.org/data/doc/gstreamer/head/gst-plugins-good/html/gst-plugins-good-plugins-splitmuxsink.html

If you use jetson_multiemdia_api, we have 01_video_encode sample for demonstrating hardware video encoding. It is low level and outputs h264 stream. You would need to implement the muxer to get mp4, ts or mkv file.

4

Thanks Danelll for quick response. However, I want to understand turning off nano without properly shutting down will have impact on video( as file won’t be closed properly) or have any impact on OS drivers

5

Hi,
If you cannot exit video encoding gracefully, you can use .mkv or .ts. The containers are designed for streaming and can keep most video content in sudden shutdown, although there is still chance you will lose several tail frames.

Not sure how you shut down the device. Suggest not to cut off power arbitrarily. This may harm the hardware module.

6

Thanks Danelll.

How can I gracefully shutdown Nano by using switch/button. Also, is it possible for me to attach one more button which can be used to start/stop video recording.

7

Hi,
Do you use gstreamer or jetson_multimedia_api? For multimedia usecases, we support the two software frameworks. You can choose one framework to implement video encoding.

8

Hi,

I am using gstreamer. But dont mind shifting to jetson_multimedia_api. All i want is start and stop the video recording process on click of button

Thanks
Karun

10

In case, if I go with mkv or ts files, is there option where I can have video generation logic separately from inference. All I want is, I want to segregate these two logics so that, recording of video won’t have any impact on inference logic. This way, I can have both these executed in parallel.

11

Hi,
You may run two pipelines in single process. One pipeline is

appsrc ! nvvideoconvert ! nvv4l2h264enc ! ...

And the other is

appsrc ! nvvideoconvert ! nvstreammux ! nvinfer! ...

There is no existing implementation. You may implement it based on this sample:

/opt/nvidia/deepstream/deepstream-5.0/sources/apps/sample_apps/deepstream-appsrc-test
12

Thanks Danelll. My requirement is, My inference program is taking 5 min for processing 1 min video. As I am not looking for real time processing, I am evaluating option of segregating both logics. This way, I can make sure that delay of inference logic won’t have any impact on recording of videos.

13

Hi Danelll,
I have been looking for sample or direction from experts on how to do delayed process of frames without impacting video recording. Would be of great help if you can share some design or python sample to proceed

14

Hi,
We have implemented some key event to control the pipeline:

There is no sample fitting your uscase. You may check how deepstream-app is implemented and do customization.

1 Like