I am from vismai analytics, a startup working on video analytics, we are primarily working on intel based machines(using OpenVINO and NCS2).
We had demonstrated the OpenVINO Face Recognition(https://docs.openvinotoolkit.org/latest/_demos_python_demos_face_recognition_demo_README.html), the results we got for frontal face recognition on Indian Faces are close to 80% and we encountered loads of unknown face labels, due to the face angle from a CCTV camera. In order to increase the accuracy, Can you please share me the process to retrain the FR model.
I had also seen that its a pytorch model, if possible, Can you please share us the pytorch model and code to retrain the model and increase the accuracy.
Hi Mounagurusamy, Guruvishnuvardan,
You can find more information about the face recognition models in the repository here. The pre-trained model included is not open sourced, so the information there is all the available info. Although the face recognition model is available in IR form only, you may find PyTorch* face recognition model in OpenVINO training extension here. Hope this answers your question.