Robot Not a Dev? Buy Now
Not a Dev?

Misty Community Forum

Facial recognition issue

Currently, facial recognition works best in a well-lit area.

Hey Donna!

This is a technical limitation of camera hardware. While there are cameras that can work at very low lighting levels, our current hardware works best in good lighting. I hope this helps!

2 Likes

I’m not getting anything when trying this. Which camera does it utilize? I don’t see the light ever come of for the Microsoft Cam. The LED does do something when I send the training command and it says

[{"result" :"succes"}]

but then for detection and recognition, nothing really happens, the data block is just {}
Any thoughts?

Face detection and recognition utilize the RGB camera (the MS Lifecam).

Currently the API returning the result : “success” means the computer vision service received the message. We are adding more detailed status messaging so you can be more informed as to what is going on low level.

The light not coming on the Microsoft Cam is an issue. The CV service is not receiving camera frames if that light is not on. Can you make sure it is plugged in and reboot the robot?

Additionally, have you ever had that blue light stay on while the robot is on?

1 Like

I have unplugged and rebooted a number of times as well as reverted/re-updated versions of the application and firmware a few times as well. Oddly In none of those scenarios did the camera light stay on other thank a blink on initial boot.

I messaged you on Slack, @station. I want to walk you through some instructions on confirming the config for the LifeCam.

1 Like

for posterity or maybe just the others who might have this issue, the way to fix it:

  • plug in the usb to micro usb into the 820 adb dev port (closest to the USB3 ports)
  • run the following shel scripts (assuming you have adb in the local path or installed on the system)
> adb shell
> cd sdcard
> ls

you should see USE_USB.txt in there
if not: touch USE_USB.txt then reboot

This fixed the issue so that the camera (Microsoft one) turns on with the blue light.
Face training, detection and recognition all worked after this.
Thanks Again @Scott ! :robot: :sunny:

3 Likes