Facial Authentication

The TeamWork Kiosk supports facial recognition to verify employees and help prevent "buddy punching".  The kiosk must run in a device that both has a camera and supports access via one of the major browsers.  This article describes the setup and use of facial recognition to authenticate employees when clocking.

(Note:  Kiosk with Facial Recognition is only available in TeamWork 5.  These screenshots are for the old UI, but the same functionality is available in TeamWork 5.)

Setup Policies

There are two mode to use facial identification and an ‘off’ setting.


Detect is a good session to use in the beginning. It will only make sure that system can detect a face in the picture and is the first step to start using Facial Identification. The detection confidence is a setting to set the sensitivity that should be used for the Detection algorithm. In our testing though, the detection was more usually close to 100 when a good picture was taken, and close to 0 if no face was in the picture. We would recommend a value no lower than 80, but we don’t think that this value would have much of an impact.


When Identify is on the system will start comparing the photo taken with the photos stored for the employee.

The identification Similarity is the threshold for how confident the system is in identifying the employee against known images. A detection above 90 is usually a good stringent threshold but might stop identifying the person as (s)he ages over a couple years. In our limited testing with 50 different faces, none got over 60% in comparing them to the other 49 images. I would not recommend going below 80 though.

When turning on detection, you will probably want to go very low on the Identification Similarity for a period and increase it over time. You might even want to start with 0 (basically the same as detection, but we start checking for similarity and store of the values of the similarity) and increase it to a value in the 80-90 range over time.

Report failed Identifications will log records that can be viewed Time Issues. This is for scenarios where an employee has problem getting identified, we have the time stamp of when they tried to clock and the picture that wasn’t approved.


Don’t forget add the Policy Component to the Policies.

Enrolling Employees

In the employee section, open an employee and go to Pictures section. This is one of the places that you can enroll pictures for an employee. Here you can either use the camera to take a photo or upload a picture from the device to enroll as photo used for Facial identification.

At this time we would encourage users to not upload more than one or maybe two pictures of an employee, as it will increase the time for a failed identification by upward of a second per picture. Having 10 pictures of an employee for the system to compare with and getting a bad picture of an employee it could take upward to 10 seconds to get a failed response back.

Time Section

Under Review in the Time section there is a menu option for Photos.

This page show the pictures that are taken of an employee once they can be used for enrolling pictures of an employee. Once you have an employee enrolled you can also check how well a picture Identify against the enrolled picture(s).

Once you have identification going the Id Similarity score should also be shown on this page. Additionally, you can also filter the page on similarity (both under and over a threshold) to be used to see what impact a change in the Id Similarity in the policy would make in not approving Facial Identification.

And with Identification turned on the similarity will be listed.

Failed face failures

The first picture taken of an employee will not be logged as failure.

In the kiosk, when the employee logs in and if that picture fail authentication they will be asked to retry, if that try also fails, that picture will be logged (in both Detect and Identify), the first will not. They will be reported as Facial Id Failed in the Issues list in the time section.

When open the Issue the confidence will be reported and if you open the photo tab, you will see the failed picture.

After reviewing the picture you should void the Issue/Exception so it ‘disappears’ from the list.