Hello,
I am working with the NUCLEO-N657X0-Q board and using the object detection example from the model zoo. In the current configuration I see that the YOLOv2 model can be deployed only with uint8 input and int8 output.
However, I noticed two post-processing functions in the code:
- od_yolov2_pp_process — for input data of type float32_t.
- od_yolov2_pp_process_int8 — for input data of type int8_t.
This leads me to think that the model might support float output. I would like to clarify:
- Is it possible to deploy the YOLOv2 model with float output?
- If yes, what changes need to be made:
- In the deploy script (e.g., configuration parameters)?
- In the project (e.g., configuration files or code)?
- In the user_config file?
Thank you for your help!
