TDA4VP-Q1: Yolov8 ONNX->TIDL conversion example

We are currently working on deploying a YOLO-like model (specifically based on YOLOv8) onto a TDA4VP-based device, for which we are required to use TIDL-RT. We’ve reviewed the available documentation and explored useful tools on GitHub (such as edgeai-tidl-tools and edgeai-tensorlab). However, some aspects of the deployment process remain unclear.

 

For instance, YOLOv8 models are mentioned in the documentation, and limited examples are provided (such a prototxt file: link). Yet, these models do not appear in model_configs.py or model_zoo, which makes it harder to test or use as a reference.

 

Thus, several questions:

1. Could you clarify why no Yolov8 examples can be found in model zoo - is this due to internal prioritization, or some technical constraints? Or maybe question of license?
2. In the prototxt file linked above, we see "head_param" values. What exactly do these refer to in the context of a Yolov8 model? Is it the same as "yolo_param" for Yolov3/v5?