about face detection and face recognition

[lai__wilson]

Hi Intel,

I have 3 question as below.

About face detection (FD) and face recognition (FR),

1. about precision for FD and FR,

    Do you know how precision for FD and FR is now?

    Do you know how to improve precision for FD and FR becasue openVINO just can register one face to data base?

    I used new openVINO sdk (2018.1.249) to build 2 scripts for testing FD and FR feature using ubuntu16.04 and kernel4.14.    

    For FD test, it still can not detect a lot of faces including face darkness/ face brignthess/ faceing to left/ faceing to right/ cuting jaw/ no glasses/ face distance for 260 cm. 

   For FR test, it will recognize wrong face from registed face in data base.

   This is one of  the most serious.

   So Do you know how to improve presision for FD and FR?

2. About camera issue,

   Currently, I want to test FD and FR for multi camera.

   But it just can test USB port in main board.

   It just can use one camera if I used USB HUB with power supply.

   This script for FR will show error message after pluging 2 camera in USB HUB.

   error message:

   Error: no input image 

   Do you know how to use USB HUB with multi camera success?

3. About camera to video,

   Currently, I can save camera frame into video with .avi success.

   And I can play this .avi success.

   But I can not heard any sound in this .avi file.

   Do you know how to record sound and video in the same .avi file?

Thanks.

[lai__wilson]

Hi Intel,

Have you any update about my questions before?

Thanks.

[Monique_J_Intel]

Hi Wilson,

I am currently looking into your questions and will respond with answers by the end of day.

Kind Regards,

Monique Jones

[Monique_J_Intel]

1.)For the face detection models we have precision FP32 and FP16 available. Are you meaning accuracy? Also, what face detection and face recognition models are you referring to?

2.)What script for Facial recognition are you referring to and can you give me more info on the command that you ran to execute the script or even attach the script?

3.)yes, you can do so with Vokoscreen.

I highly recommend that you upgrade to the latest version of OpenVINO that has improved face detection models and samples like the multi-channel sample and interactive face detection demo.

Kind Regards,

Monique Jones

[lai__wilson]

Hi Monique,

1. Yes, I have some questions as below.

    Do you know what accuracy value of FD and FR is using openVINO in your test?

    And Could I improve its accuracy of FD and FR becaseu openVINO just can register one face into data base? 

    Do you know the distance of FD/FR of camera (in real time) why it is 260cm?

    And Do you know how to improve the distance of FD/FR of camera? 

2.  I used FR script of cvsdk_sample_facerecognizer under /opt/intel/computer_vision_sdk_2018_1_249/opencv/share/Opencv/sample/.

     And its source code is facerecognizer.cpp under pvl/.

     Is it correct for verifing FR?

3. Sorry, We don't want to use Vokoscreen.

    We want to use openVINO to record video and sound into one .avi file uisng camera in real time.

    Do you know openVINO why it can not record any sound?

    or Do you know openVINO how to record sound and video into one .avi file for our customer?

Thanks.

[lai__wilson]

Hi Monique Jones,

     I used FR script (cvsdk_sample_facerecognizer) under /opt/intel/computer_vision_sdk_2018_1_249/opencv/share/Opencv/sample/pvl/ for verfing FD/FR of image and camera.

     Do you know what accuracy value and FPS for this FR script is for image and camera?

     I used openVINO sdk with 2018.1.249v.

   Thanks.