Setup:
• Hardware Platform: Jetson
• DeepStream Version 6.4
• JetPack Version (valid for Jetson only)
• TensorRT Version: R35 4.1
I have a deepstream pipeline that has an object detector as PGIE, and a regression model as SGIE. I have based this off the dstest2 example.
I know that the PGIE is working as expected, but the issue is that the SGIE is not outputting correct values.
The PGIE has
output-tensor-meta=0
The SGIE has the following configured values:
network-type=100
output-tensor-meta=1
infer-dims=3;224;224
operate-on-gie-id=1
operate-on-class-ids=1;2;3;4
As well as model specific configuration.
I have attached a probe to the src of the SGIE, and am using this to access the output of the SGIE.
I am accessing this data at:
batch_meta->obj_meta->user_meta->user_meta_data->output_layers_info[0].buffer
Using a debugger I can confirm that the output_layers_info is coming from the SGIE, which has 1 output layer named “length”.
However, this keeps outputting a value of 0.
I have tested using this model as the PGIE, and when I do so I see the expected output (which is ~200), but I have not been able to get any output besides 0 when this model is used as the SGIE.
I have swapped out the models and configs for the ones from dstest2 and confirmed that when I am using the dstest2 PGIE and SGIE I am able to view non-zero data at batch_meta->obj_meta->user_meta->user_meta_data->output_layers_info[0].buffer
How can I resolve this or determine why it is happening?
My aim is to get the correct value out of SGIE on the SGIE src probe.