Request for the proper documentation for converting yolo nas custom trained model to openvino format(.xml and .bin)

[kenilaivid]

🚀 Feature Request

I am writing to kindly request your assistance with a feature that would greatly benefit the community using Super-Gradients and OpenVINO. Specifically, I am seeking proper documentation on converting YOLO NAS custom trained models to the OpenVINO format, specifically the .xml and .bin files.

The YOLO NAS (You Only Look Once Neural Architecture Search) model has gained significant popularity due to its high accuracy and efficiency in object detection tasks. Many developers, including myself, have trained custom models using the YOLO NAS architecture to suit specific use cases. However, when it comes to deploying these models in production environments, it becomes crucial to convert them to formats compatible with popular inference frameworks like OpenVINO.

OpenVINO, with its support for a wide range of hardware platforms and optimizations, is an excellent choice for accelerating deep learning models. Unfortunately, there is a lack of comprehensive documentation on the specific steps and processes involved in converting YOLO NAS custom trained models to the OpenVINO format. This documentation gap makes it challenging for developers like me to seamlessly integrate our models into the OpenVINO workflow.

Hence, I kindly request your assistance in providing proper documentation that covers the entire conversion process, specifically targeting YOLO NAS models. The documentation should include clear instructions, step-by-step guidance, and any necessary code snippets or scripts required to convert the model's architecture and weights to the OpenVINO format, producing the essential .xml and .bin files.

Having access to detailed documentation will not only benefit me personally but also help numerous developers who are using Super-Gradients and OpenVINO for their computer vision projects. It will enable us to leverage the power of OpenVINO's hardware acceleration and optimizations, leading to faster and more efficient inferencing of our YOLO NAS custom models.

Looking forward to your positive response and the possibility of seeing this valuable documentation become available in the near future. Please let me know if you require any additional information or if there is anything else I can do to assist in the process.

Proposed Solution (Optional)

No response

[BloodAxe]

If someone from the community is willing to participate in studying this topic and contributing a tutorial for exporting YOLO-NAS to OpenVINO feel free to reach out.

[Y-T-G]

I have implemented YOLO-NAS in C++ with OpenVINO backend.