Intel® Distribution of OpenVINO™ Toolkit
Community support and discussions about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all things computer vision-related on Intel® platforms.

Retail Face Detection model with face masks

ET-Foley
Beginner
505 Views

Hello Community,

do you have experience with the face detector and the current face masks in the retail environment?

Is it planned to train it with face masks again?

Thank you very much for your feedback.

Greetings,

ET-Foley

0 Kudos
9 Replies
Munesh_Intel
Moderator
483 Views

Hi Thomas,

Thanks for reaching out to us.

We don’t have any face detection model with face masks currently.

As for future enhancements with face masks, with due apologies, we regret to inform that we are unable to comment on future support as well.

Alternatively, I would suggest you have a look at the following GitHub link, 'Face Mask Detection using OpenVINO'.

https://github.com/mmphego/face_mask_detection_openvino


Please be noted that all the materials available in the above link is not related to Intel and we cannot guarantee anything about them.


Regards,

Munesh


ET-Foley
Beginner
473 Views

Hi Munesh,

thank you so much for your fast reply!

We are currently working with a larger team to train more data sets for our own applications.

Since the OpenVINO Retail Face Detector works great, we do not want to build our own training data set with faces.

However, we currently have several thousand faces with face masks, which would make the detector much more stable in detection.

Could we possibly work together so that we can send you a part of the labelled data and you can re-train your model for us and the community?

In the end, everyone would benefit and the algorithm.

 

Best regards,

ET-Foley

Munesh_Intel
Moderator
460 Views

Hi Thomas,


I've conveyed your request to the relevant team. Let's wait for their feedback.


Regards,

Munesh


Munesh_Intel
Moderator
443 Views

Hi Thomas,

Thanks for waiting. The feedback we’ve got is that Intel is unable to support your request due to Intel’s policy towards publicly available data.


Having said that, I would suggest you collaborate with members of our Community for the enhancement of the existing face detector models with face mask, or in the development of new face mask detectors.


Here is a tutorial to help you get started:

https://www.pyimagesearch.com/2020/05/04/covid-19-face-mask-detector-with-opencv-keras-tensorflow-an...


Additionally, do check out our public model, RetinaFace-Anti-Cov, which is a customized one stage face detector to help people protect themselves from COVID-19.


GitHub link:

https://github.com/deepinsight/insightface/tree/master/detection/RetinaFaceAntiCov



Regards,

Munesh


ET-Foley
Beginner
426 Views

Hi Munesh,

It is a great pity that we cannot cooperate with each other here. Developing a Face-Mask Detecktor is not really a problem for us. It's more about the face detector which would have to be improved for the detection of faces with face masks.

Because of our hardware we are dependent on the face detector for retail. Is there a possibility to train it further?

https://docs.openvinotoolkit.org/2021.1/omz_models_intel_face_detection_retail_0005_description_face...

Thanks for your support!

 

Best Regards,

ET-Foley

Munesh_Intel
Moderator
379 Views

Hi Thomas,

We've a good news for you. We've got feedback from Open Model Zoo developers that if you are agreeable to publishing your dataset under a permissive license, they would consider using your dataset to train a model.


Regards,

Munesh


ET-Foley
Beginner
375 Views

Hi Munesh,

that sounds excellent!


Can you put me in touch with your colleague?

Can you maybe send me a PM so that I can send you my contact details?

 

Best regards,

Thomas

Max_L_Intel
Moderator
311 Views

Hi Thomas,

We are going to close this community thread. We'll continue to maintain communication through email.

Thanks.
Best regards, Max.

Munesh_Intel
Moderator
363 Views

Hi Thomas,


I've sent an email to you. Let's bring it forward from there.


Regards,

Munesh


Reply