Deepstream with RTSP vs files

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

• Hardware Platform (Jetson / GPU) : NVIDIA GeForce RTX 3090
• DeepStream Version : 6.3
• JetPack Version (valid for Jetson only)
• TensorRT Version : 12.2
• NVIDIA GPU Driver Version (valid for GPU only) : 535.104.05
• Issue Type( questions, new requirements, bugs)
• How to reproduce the issue ? (This is for bugs. Including which sample app is using, the configuration files content, the command line used and other details for reproducing)

I am using deepstream-test3.py and added tracker and other models

I have 32 video files when run pipeline with these video files get objects count from 400 to 600 but when use tool to simulate these video to RTSP and pipeline we get objects count from 100 to 250
also we don’t set parameter batched-push-timeout
why these drop happened however used same videos

3

Could you print the total number of frames in those 2 scenarios?
You can also use just 1 video to compare.

4

I did test you mentioned on wild track video number 1 on file and also simulated as RTSP

Screenshot for test on video file

video_file
video_file1928×1038 184 KB

Screenshot for test simulate video file as RTSP stream

rtsp_stream
rtsp_stream1928×1038 187 KB

deepstream-test3 script we used to do this test
deepstream_test_3.txt (18.1 KB)

as you can see count of objects in video file greater than count of objects in RTSP stream however we are using same video so count should be the same

why this issue happened?

Appreciate your feedback

5

There may be recoding or packet loss when you use the RTSP.
1.You need to make sure that the video is not recoded when you use the rtsp.
2.You need to make sure there are no packet loss when you use rtsp.

6

I am using cvlc tool to simulate RTSP from videos on same server that has deepstream_test.py script so I think no network drops

or what do you think?

7

It cound lead to the two problems I mentioned above with the method you used.
1.Vlc may decode the stream first, then re-encode it, and then send it though the rtsp.
2. DeepStream is based on Gstreamer. The rtsp module in gstreamer may lose packets while transmitting the video data.

8

do you recommend other RTSP simulation tool to get same count of objects from RTSP as in files?

My goal to simulate RTSP and get same results as I get from files

Please help

9

Could you refer to Build rtsp server to build the rtpserver? Could you attach your probe function and where did you add the probe function?

10

I tested using video files , RTSP simulation using cvlc and ffmpeg

below screenshot clvc to simulate RTSP

rtsp_stream_by_cvlc
rtsp_stream_by_cvlc1928×1038 205 KB

below screenshot ffmpeg to simulate RTSP ( Gst-nvstreammux:)

rtsp_stream
rtsp_stream1928×1038 187 KB

below screenshot for video files

video_file
video_file1928×1038 184 KB

I tried also ffmpeg tool as you see and still count of objects in files higher than ffmpeg and cvlc

Can you elaborate on this results and do we need to do any settings to get same results for RTSP as in video files?

11

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

Let’s narrow down that first.
1.Could you try to use the tcp protocol as lower transport protocols of the rtsp?
2.Could You use a h264 stream with just 1 frame for comparison?

12

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