In the config file of DS, is parse-bbox-func-name=xxxx fixed? What if the output is not bbox, can we call the setting via other names? Like parse-customName-func-name=xxx?
Deepstream support parse network type detector and classifier, what’s your network model type?
It’s openpose-plus. Output is heatmap and paf. We would like to draw the skeleton using gst-nvdsosd.
Refer to this topic, similar issue, https://devtalk.nvidia.com/default/topic/1071353/deepstream-sdk/does-nvinfer-work-with-custom-neural-network-/post/5427791/#5427791
Could u share how to use openpose with deepstream? I use ./exectrt to convert the openpose caffemodel and prototxt to an engine file and I use it in the config file.
But in the custom function, when i compare the netoutput data with the result of inference with opencv, its different. I cant get the correct Keypoint.
Please help to open a new topic for your issue. Thanks