setting up cuvid decoder for h264 decoding with rtp stream data

Accelerated Computing CUDA CUDA Programming and Performance
#1

Hi,

I successully integrated the codec sdk into my application, and now want to feed the decoder with live data coming from an rtp stream.
Is there any information on how to setup the decoder properly (I have the sdp information for the stream). I can feed the rtp payload (which is plain h264) into the parser/decoder, cuvidParseVideoData returns CUDA_SUCCESS, but there are no frame coming out from the decoder.
How do I use the sdp information to fill the CUVIDEOFORMAT struct properly, I think that this is the problem.

Thanks for any help or information

Andi

#2

Hi,

I was able to fix the problem, found a solution myself. There were a few things to do.

  • I not only parsed the sdp information from the stream, I also injected the SPS/PPS data into the data fed to the parser.
  • RTP data was segmented, I had to create full NAL units before feeding the parser
  • data contained in the RTP stream did not have NAL unit markers, I added them myself

    • After all, this works. Even with multiple streams at once.

    Regards
    Andi