Why Facebook’s new research validates D-ID’s Smart Anonymization solution
I smiled the other day when I read on TheVerge that Facebook researchers now claim they are able train AI to fool facial recognition systems. We at D-ID have been doing this successfully for almost two years. In fact our own video anonymization solution was launched just over two months ago. Still, it’s always nice when an established market player validates your technology and business approach.
So to all our colleagues at Facebook’s AI Research team, on behalf of all of us at D-ID: Kudos! And welcome!
According to Facebook “…recent world events concerning the advances in, and abuse of face recognition technology invoke the need to understand methods that successfully deal with de-identification.” We completely agree. In fact, our very name – D-ID – stands for De-Identification.
Our approach is called Smart Anonymization and it works by removing facial images from videos and still images, and replacing them with computer-generated, photorealistic faces of nonexistent people. The anonymized faces preserve the key, non-identifying attributes of the original faces, i.e. age, gender, expression, gaze direction, motion etc., but remove all other personally identifying information.
Here’s an example of how we anonymized a video of former British Prime-minister David Cameron (watch it on YouTube):
Smart Anonymization helps protect people’s privacy – especially in video footage being shared online (even in live streamed video). But it also has multiple commercial use cases whose potential is still mostly untapped.
Valuable Data is Going to Waste
A white paper released by Cisco last year claims that video-data alone is expected to make up 82 percent of all Internet traffic in 2022. Video traffic today is generated mostly by users – in 2017 YouTubers were uploading nearly 300 hours of video every minute – a number that has since grown; Facebook today claims to see 100 million hours of daily video watch time. These numbers are expected to grow exponentially over the next 2-3 years, fueled by non-consumer services like smart city and retail applications, autonomous vehicles, and other sources that generate petabytes of data every minute.
Yet – while video streaming/sharing data figures are impressive, only a fraction of it is being used commercially for gathering information and for data analytics. This is due to the fact that much of the visual data today contains sensitive Personally Identifiable Information (PII) such as human facial images. Privacy regulations like the European GDPR, the California Consumer Privacy Act (CCPA), the Illinois Biometric Privacy Act (BIPA) and others, restrict the use of visual data and do not allow it to be used for harvesting business intelligence data.
Smart Anonymization solutions, like D-ID’s, are about to change that.
Above: where traditional methods fail, Smart Anonymization succeeds
The Solution: Computer Generated Faces
The real advantage of D-ID’s Smart Anonymization approach is that it ensures full compliance with privacy regulations, while at the same time enabling businesses and individuals to benefit from full and unhindered use of digital images. So in essence, we’re creating new business opportunities with a true win-win solution.
And yes. We’re extremely happy that Facebook now shares our point of view.