Hi,
I have evaluated DeepStream 4.0 EA on Xavier,
and successfully launched test programs.
The performance was amazing!
Then, I have modified deepstream_test1 as test5 as the link below,
so that it could accept an RTSP stream instead of a raw H264 file.
It works, but it looks like the primary detector takes longer exponentially,
according to messages as follows coming from osd probe.
Frame Number = 20 Number of objects = 0 Vehicle Count = 0 Person Count = 0
Frame Number = 21
...(0.5 secs)
Frame Number = 22 Number of objects = 0 Vehicle Count = 0 Person Count = 0
Frame Number = 23
...
Frame Number = 62 Number of objects = 0 Vehicle Count = 0 Person Count = 0
Frame Number = 63
...(3 secs)
Frame Number = 64 Number of objects = 0 Vehicle Count = 0 Person Count = 0
Frame Number = 65
...
Frame Number = 124 Number of objects = 0 Vehicle Count = 0 Person Count = 0
Frame Number = 125
...(8 secs)
Frame Number = 126 Number of objects = 0 Vehicle Count = 0 Person Count = 0
Frame Number = 127
...
Frame Number = 188 Number of objects = 0 Vehicle Count = 0 Person Count = 0
Frame Number = 189
...(15 secs)
Frame Number = 190 Number of objects = 0 Vehicle Count = 0 Person Count = 0
Frame Number = 191
...
By increasing the latency property of rtspsrc,
it plays quickly in the beginning,
and then, getting slow down as described above.
So, this doesn’t look like a simple performance issue.
I would appreciate if anyone could give me a clue.
Also, in such a case, is there anyway I could debug the flow in a pipeline?
Thanks,
Takenori