Deepstream-test2 Internal data stream error

I thought you use default builtin pgie model which is a detector model.

@amycao Sorry for the confustion.

I am attempting to run deepstream-test2 application but modified to use my own classification model(s) for pgie and sgie. When I try to run this I get the error from the original post.

@amycao To my knowledge there are two example codes for how to use sgie element: deepstream-test2 and back-to-back detector from github.

The first example code uses detector (pgie) ==> classifier (sgie).

The second example uses detector (pgie) ==> detector (sgie).

I am attempting to do classifier (pgie) ==> classifier (sgie), but I get the above error.

Can you suggest how I can resolve the error and run this type of configuration?

It’s fine to do classifier (pgie) ==> classifier (sgie)
Did you use nveglglessink as output? or something else?
if nveglglessink, did you run from desktop or another way did you export DISPLAY before run the sample?

No I receive the same error above even when replacing with fakesink.

I try to mimic your test case, by modifying test2 sample, using builtin carcolor model, carcolor model(pgie) → carcolor model(sgie), without error you met, can you share how to repro your issue?

Thank you I will try your experiment and share the reuslts.

@amycao Are you able to share the inference config file you used for PGIE and SGIE in your experiment? Thank you.

dstest2_sgie1_config.txt (3.5 KB) dstest2_sgie1_config-1.txt (3.5 KB)
first one is for primary gie, second is for second gie.