When converting a Yolov8 model to a HAR file and setting the recommended end nodes to be the convolutional layers at the end of the detection head, I notice that it includes the corresponding output head as part of the contents of the HAR file. However it doesn’t seem to do anything with it.
This is the structure from the yolov8n.postprocess.onnx that I am referring to inside of my HAR file.
Is this going to be compiled and used automatically when compiling to a HEF or am I going to have to manually write a post process method to account for the missing output head? If so I see that Hailo seems to have used Yolov8 specifically in their examples but I have yet to find any source code directly including and handling this issue in Tappas or the application examples. Which I find off since yolov8m should be included in the Tappas examples based on a note in the model zoo.