There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks
What do you mean about the config file? Does it mean the gst-nvinfer configuration file? If it is so, the parameters in the configuration file are decided by the model. Please make sure you know all necessary parameters listed in DeepStream SDK FAQ - Intelligent Video Analytics / DeepStream SDK - NVIDIA Developer Forums.
The “detectors” means the models which outputs bboxes.
Seem head_pose will not output bboxes. The head_pose model can be used as secondary GIE after PGIE(E.G. face detection model). According to DeepStream’s definition, head_pose is the “others” model. The “network-type” of gst-nvinfer should be 100(which means others), “output-tensor-meta” should be enabled and the model output parsing should be customized in the nvinfer src pad probe function. We have a “faciallandmarks” model sample, which is the similar case as your head_pose model. You can refer to the sample deepstream_tao_apps/apps/tao_others/deepstream-faciallandmark-app at master · NVIDIA-AI-IOT/deepstream_tao_apps · GitHub