nvv4l2decoder generates more delay than h264_cuvid when decoding

Accelerated Computing Intelligent Video Analytics DeepStream SDK
#1

Hi I’m trying to benchmark both approach using ffmpeg decoding via h264_cuvid and using a gstreamer pipeline using nvv4l2decoder.

Using ffmpeg I’m getting frames with really low delay between camera and presentation.

On the other side with deepstream sdk and it’s decoder I am getting better performance but worse delay time

This is my gstreamer pipeline that opens in a openCv VideoCapture.

rtspsrc location=“rtsp://10.50.84.110/defaultPrimary?streamType=u” user-id=admin user-pw=admin ! rtph264depay ! h264parse ! nvv4l2decoder ! nvvideoconvert ! video/x-raw , format=RGBA, width=800,height=600 ! videoconvert ! video/x-raw, format=BGR ! appsink

Is there any parameter to add to the decoder to improbe the delay time??

#2

Hi,
Please try the setting ‘appsink async=false sync=false’

#3

Thanks!! With appsink sync=false the delay is similar as ffmpeg (h264_cuvid)
I cannot put async=false. Maybe is a problem with OPENCV caps.

#4