Thank you for reaching out!
As it was stated YOLOv5 is currently not an officially supported topology by the OpenVINO toolkit. Yet you can import and convert it in DL Workbench. Then, you can infer the model and see the performance metrics (FPS & latency), profile and find the best combination of the streams and batches. As YOLOv5 is not officially supported by the OpenVINO toolkit, only the default optimization algorithm can be used for the INT8 calibration. Accuracy evaluation is not supported at the moment.
DL Workbench QA
This thread will no longer be monitored since we have provided answers. If you need any additional information from Intel, please submit a new question.