Smart classroom demo not recongizing face

[GMath7]

Hi,

I tried executing smart classroom demo application as mentioned in the webpage.

./smart_classroom_demo -m_act /home/myPC/computer_vision_sdk_2018.4.420/deployment_tools/intel_models/person-detection-action-recognition-0003/FP32/person-detection-action-recognition-0003.xml -m_fd /home/myPC/computer_vision_sdk_2018.4.420/deployment_tools/intel_models/face-detection-adas-0001/FP32/face-detection-adas-0001.xml -m_reid /home/myPC/computer_vision_sdk_2018.4.420/deployment_tools/intel_models/face-reidentification-retail-0071/FP32/face-reidentification-retail-0071.xml -fg faces_gallery.json -i cam.

I have executed create_list.py to get faces_galley.json file, after giving cropped faces . But on executing the above command in terminal, the application labels face as Unknown along with the detected activity. For those people whose faces are not listed in faces_gallery.json, only their activity is labelled with no ' Unknown' also marked.

Could you please help in sorting out the problem. Also adding the faces_galley json contents below

{"amruth": ["/home/myPC/Desktop/Faces/amruth.1.png", "/home/myPC/Desktop/Faces/amruth.2.png", "/home/myPC/Desktop/Faces/amruth.3.png", "/home/myPC/Desktop/Faces/amruth.4.png", "/home/myPC/Desktop/Faces/amruth.5.png", "/home/myPC/Desktop/Faces/amruth.6.png"]}

[Shubha_R_Intel]

Dear Gina,  OpenVino 2019 R1 was just released today. Can you download it and try ? Please post the results here.  

Thanks for using OpenVino !

Shubha

[GMath7]

Hi Shubha,

We checked the same in OpenVINO R1 2019. Still the same issue. But executing activity recognition model is fast when compared to 2018 R5 version.

regards,

Gina

[Shubha_R_Intel]

Dearest Mathew, Gina,

This may in fact be a bug. Please send me the items required to run the smart classroom demo in a *.zip file. I have PM'd you - simply respond to the PM.

Thanks for your patience !

Shubha

[Shubha_R_Intel]

Dearest Mathew, Gina,

Thank you for sending me your tightly-cropped faces and faces_gallery.json in a *.zip over PM. Now I have to ask two questions:

1) Does the video (it seems you passed in cam for webcam) you passed into -i actually represent a classroom setting ? Because unless you are facing an actual classroom with the webcam,  it will not work. Remember the models have been trained for a classroom setting.

2) do the faces which appear in faces_gallery.json appear in the classroom setting ?

The smart_classroom_demo README actually states this here:

The demo shows an example of joint usage of several neural networks to detect three basic actions (sitting, standing, raising hand) and recognize people by faces in the classroom environment.

I answered a similar forum question below:

https://software.intel.com/en-us/forums/computer-vision/topic/805529

If you are actually webcamming to a classroom setting and the faces in faces_gallery.json are in that classroom, then yes - this is a bug.

Please let me know how your situation fits.

Thanks for using OpenVino !

Shubha