I used an IP camera to detect objects based on deep learning by Jetson Nano.To improve performance, Deepstream 5.0 was used. However, at an irregular moment after entering the command, a memory problem caused the program to shut down abnormally.
So far, two errors have been found as shown in the picture above. The IP camera is connected to the Jetson nano by RTSP, and the environment is as follows.
ubuntu : 18.04
jetpack : 4.4.1
deepstream : 5.0
IP camera name : FW7902-FTF
Previously, I developed the program from Deepstream 4.0.1 to QT, It was used in h.264 format. Due to the problem of the camera delay, it was changed to mjpeg format and used. I used Deepstream 5.0 refer to the following article.
However, your answer this time was that the mjpeg format cannot be used in Deepstream. I checked that camera streaming was working on the destream-app sample, but there was a problem that the program ended irregularly. I want know if the support for mjpeg format is incomplete or not fully available.
And
mjpeg : 300ms
h.264 : 600ms
There is a delay I wonder if I can reduce the delay of h.264
Thank you.
deepstream is a SDK, it can support mjpeg rtsp. deepstream-app is a sample application, it dose not support mjpeg rtsp because we have not implement this part in the sample application. You are using uri source but not rtsp source, right? It is better to write your specific application to use mjpeg rtsp.
You’re right. I used the URI source and not the RTSP source. Then I want to know if the reason why I was able to stream the camera was because I used URI source without using RTSP source. And in my case, I would like to know the difference between using RTSP source and URI source.
Thank you for your quick and accurate response.
With rtsp source we will connect to nvv4l2decoder and set the related properties explicitly, but with uri source, the decoder and other elements are choosen by uridecodebin automatically, the properties are all using default values.
