How to use tao pretrained models with Composer?

•Hardware Platform (Jetson / GPU) GPU (T4)
•DeepStream Version
•NVIDIA GPU Driver Version (valid for GPU only) 515
•Issue Type( questions, new requirements, bugs) Question

I want to use one of the TAO pretrained models (TrafficCamNet) (from tao_pretrained_models) together with the graph Composer.

I thought I could just set the required config files in the NvDsInferVideo node, but it seems like some properties are missing, i.e. I don’t see anywhere to put the lables.txt or the trafficcamnet_int8.txt file.

It also seems that the nvinfer_config.txt doesn’t seem to be the file that the config-file-path property expects, as the graph crashes with the following error:

Failed to load config file: Key file does not start with a group
ERROR: <gst_nvinfer_parse_config_file:1303>: failed

What kind of config file is expected here? As far as I could tell, all the reference graphs use the configurations nodes from the NvDsSampleModulesExt… And the only tutorials I could find that explain how to use TAO models use the reference deepstream-app built in C++, but not a graph…


Seems I was looking in the wrong folder for the config file.
I found the correct yaml config files in GitHub - NVIDIA-AI-IOT/deepstream_reference_apps: Samples for TensorRT/Deepstream for Tesla & Jetson

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