Privacy and Facial Recognition: How to Respect Them on a Video Call

Privacy and facial recognition using biometric data are currently among the most debated topics. Whether in a public place or at work, we have to deal with devices able to detect our face, to track it, and then to compare their results with others stored in digital archives.In addition to video surveillance cameras,  other devices, such as smart glasses, are available that in certain situations can also track people’s physical, physiological and even the behavioral characteristics. Consequently, a person can end up taking part unintentionally in a live work session. Therefore, how can one make sure to respect privacy and facial recognition during video calls? With Eye4Task. Let’s discover why and how.

Privacy and Video Call with Smart Glasses: Are They Compatible?

Taking all the necessary measures to protect personal data is the answer. This can only happen by usinghi-tech solutions that improve processes. In airports, railway stations, squares and sensitive places, devices that can detect people’s physical characteristics are on the rise. In a few steps, you can univocally identify any user.
Let’s think of an expert in the field who is working in a place frequented by many people. He is wearing wearable devices. He is remotely connected to another professional. Together, they follow the steps to finish an inspection be it, for example, maintenance, logistics or even security. Although they are careful to film only those elements and aspects that are useful for their intervention, the eye of the display may frame people passing by.

How to protect their privacy? By complying with the GDPR on the treatment of biometric data and designing solutions that won’t negatively impact on the privacy of people caught by the latest generation devices, as well as informing passers-by, both users and non-users, with special forms for their consent to data processing.

Facial Detection And Face Obscuration In E4T

The collaborative platform Eye4Task has introduced a service for the detection of faces and their obscuration. To this aim, in 2020 the HeadApp experts launched an experimental version of E4T equipped with an extension that protects people’s privacy.

The experimentation looked at different pattern recognition models to identify the one best suited to product needs. In fact, the working contexts in which the operator in the field has to perform his tasks can be affected by aspects that influence facial recognition. Moreover, the pattern used must also take into account the partial coverage of the face.

Thus, initially recognition models were applied to the stream by exploiting OpenCV technology. Face detection with Haar cascades is based on machine learning where a cascaded function is trained with a set of input data. “This approach,” HeadApp experts explain, “did not yield the expected results in terms of performance. This is because of the delay in applying the filter on a live stream from mobile devices.”

How E4T Respects Data and Privacy during Live Broadcasts

The experimentation then led to the integration of the recognition models directly on the image source device. Then, the pre-trained models were integrated through the image classification of TensorFlow in its Lite version.

After positive feedback, the team moved on to the next phase: stressing the models in order to receive more timely and responsive feeds, even in case of partial coverage and a high number of subjects to be identified at the same time.

“We identified the solution in BlazeFace. It’s a lightweight, high-performance face detector designed for mobile GPU inference. It runs at a speed of 200-1000 + FPS on flagship devices. This real-time performance allows it to be applied to any Augmented Reality pipeline that requires an accurate region of facial interest.”

With E4T, then, face detection is performed and then faces are obscured already during the video call with the Support Room. A further step towards respecting privacy and personal data occurs in the video capture both in public places and in sensitive contexts, such as research centers, military bases or places with company secrets and classified information, for security checks.

For more information on the HeadApp software, contact us for a free consultation or visit our Solutions page.

Request a Demo!