TensorRT engine giving wrong/different output in DeepStream

@y14uc339 @hidayat.rhman Any luck with using the tensorrtx repo with deepstream? I too am finding it works fine in TRT, but with deepstream the ‘prob’ data passed to the custom parser does not match the data in TRT.

1 Like

@kbass Nope

+1 Need help!

Hi, am wondering if you fixed the problem? Have the same issue.

Hey you need to mention this line at then end of yololayer.h
REGISTER_TENSORRT_PLUGIN(YoloPluginCreator)

I solved this by following this Fix cuda synchronization

The repo’s owner only do synchronization to yolo in tensorrtx so the problem occurs while I trying to deploy retinaface on deepstrem-6.0. Fix synchronization in decode.cu file will solve this problem.

1 Like

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