Interview: Cognitec’s Elke Oberg on real-time face recognition
10 December 2015 15:28 GMT

Face recognition in real-time has been a long-running challenge for the industry, with facial movement and stable network links among the many factors making this key surveillance task difficult to master.

Typically, the systems also require the capability to estimate age estimation and gender detection, further increasing requirements.

However, the accelerating proliferation of “smart city” projects across the globe is likely to make real-time identification an increasingly pressing issue.

Planet Biometrics talked to Cognitec’s ‎marketing manager, Elke Oberg, about the complexities of the technology and solutions involved.

What are the key challenges in creating real-time solutions for facial recognition?

Face recognition integrations are usually confronted with the existing infrastructures and equipment end users plan to repurpose. Challenges start with low bandwidth of existing networks, use of low-resolution cameras and inadequate lighting situations. Existing cameras are often placed too high, resulting in angles that considerably reduce the image quality and therefore the accuracy of the technology. In many cases, the design and layout of the facility prohibit the installation of cameras in optimal locations.

Aside from such practical obstacles, real-time face recognition applications will never reach the level of accuracy compared to image matching applications for portraits acquired in a controlled setting. Such expectations should be addressed honestly and constructively at the onset of any real-time face recognition project.

What trends have developers needed to work on to keep enhancing your solutions?

Algorithm research and development in recent years has concentrated on improving the recognition performance on difficult image material, in particular facial images with non-frontal poses or strong illumination artifacts, both typical for real-time video screening applications. And in March 2015, Cognitec introduced a highly specialized IP video camera with built-in face detection and tracking technology. The camera provides optimal image quality for real-time face recognition, even under challenging conditions, while requiring low computing hardware and bandwidth resources.

Are age estimation and gender detection something which can be applied to mobile, real-time solutions?

Yes, our technology can provide real-time gender and age range data based on anonymous facial analysis, and the results could trigger an event on a mobile device. As mentioned above, the accuracy in a real-time setting highly depends on camera quality and positioning, as well as sufficient illumination of the scene. Therefore, it is still difficult to use the technology for the reliable delivery of personalized advertising or other targeted promotional campaigns. 

Are image comparisons to large databases possible in real time on mobiles?

For investigation use cases, a mobile device can be used to take a photo at the scene, and then submit it for comparison to a central database. The central application will quickly return a result or a candidate list. Storage capacities on mobile devices continue to increase, so storing a sizable database on the device itself will soon be a plausible scenario.

Our video screening technology can send a real-time notification to a mobile device after a person was seen by a camera and matched with an image in the database.

Do you see Cognitec solutions diverging into the iOT and other new sectors?

Innovative applications for face recognition technologies are emerging every year, as we receive requests for our technology from companies developing new software and hardware products. For use cases around the Internet of Things, face recognition vendors in particular need to promote the responsible use of such technologies. The public would quickly reject biometric technologies if they are in any way abused for connecting a person’s identity and various data sets without their consent.

What do you feel is the best outreach strategy to convince the public of face recognition’s many positives?

By encouraging people to use it! If face recognition is featured on your phone or computer, use it to replace your passwords. If an airport has installed eGates for automated border control, use your biometric passport and enjoy the ease and speed of self-service entry or exit procedures. If your bank enhances authentication procedures with face recognition and other biometrics, benefit from the convenience and better data protection. It will be interesting to see which applications will receive the public’s approval and secure their/our future in this world. 

Related articles

Cognitec adds iOS support to face recognition algorithm
26/09/14
Cognitec Algorithm comes out on top in NIST Independent Vendor Test
07/04/14
Cognitec brings specialized face recognition to the border
20/12/13