Setting up my own graph in graph composer doesn't work, however the reference graphs do work

[bram.kostermans]

• Hardware Platform: NVIDIA GeForce RTX 2060
• DeepStream Version: 6.2
• TensorRT Version: 8.5.2.2
• NVIDIA GPU Driver Version: 525.105.17

I’m testing the graph composer, and while the reference graphs are working, I don’t understand that when I’m creating similar graphs using the composer (instead of loading them from file), I get errors like:

2023-04-26 16:03:00 [4,075,353ms] [Error] [omni.kit.app._impl] [py stderr]: 2023-04-26 18:03:00.315 ERROR gxf/std/type_registry.cpp@48: Unknown type: nvidia::deepstream::NvDsStreamMux:

However I am able to run the reference graph: deepstream-test1.yaml

In the yaml files I can see that for example the versions are different, however I am not able to change thos from the interface.

demo.yaml (3.9 KB)
deepstream-test1.yaml (5.1 KB)

Do you know what I should do to create my own graphs?

[Fiona.Chen]

The inferencing model configuration is missing in your demo.yaml.

Your demo.yaml can work after I added “config-file-path: /opt/nvidia/deepstream/deepstream/samples/configs/deepstream-app/config_infer_primary.txt” parameter to the “Video Inference27” extension parameters.

demo.yaml (4.1 KB)