Unable to apply object detection on rtsp stream after rtsp support update

Hi, I cloned the new repo from jetson.inference and i am unable to stream rtsp using the new update videoSource and videoOutput

This is the code I used

This is my error code

Thanks for your help

The error failed to discover stream info means that GStreamer was unable to connect and automatically determine your RTSP stream’s properties (such as the codec).

When you launch the program, try setting the --input-codec=<codec>, for example your RTSP stream is encoded with H.264, use --input-codec=h264:

camera = jetson.utils.videoSource("rtsp://", argv=['--input-codec=h264'])

Also I see that your URL is a .sdp file, I haven’t personally tried that before with GStreamer. If you continue to have issues, you might want to try a standalone GStreamer pipeline first (i.e. launched with gst-launch-1.0) to see if it’s able to successfully connect to your RTSP source.


I managed to get it working but the delay between the rtsp camera and the display on the monitor is too long, is there any way i can reduce this delay? Like the argument latency=0?

You can play around with the latency argument by uncommenting this line and re-compiling:

After editing it, run make and sudo make install from your jetson-inference/build directory.

If you try different latency settings and it improves the latency, please let me know so I can enable it in master. I am unable to reproduce the latency from here.

Also, you may try running your RTSP camera at a lower resolution, which should help reduce the latency too.

Thanks dusty, it is working fine now. Is there anyway to uncap the framerate because I noticed it is capped around 20 or is it limited to the camera i am using?