The agitation in May 2008 was immense. Whether it concerns sports event videos or digital photos of populated streets - if persons can be identified on them, they are considered to be visual data in compliance with the new EC-Data Protection Regulation (DSGVO) that has come into effect and they are therefore to be protected. This is good news for those citizens who are worried about their privacy. However, it is feared that this regulation will have a negative effect on almost anybody carrying a digital camera. In respect of sports-, wedding-, press-, or street photographers this could well mean the end of their professional career. Also, companies from other industries would be affected in as far as collecting and processing visual data is critical for their business.
New Facial Expressions due to Artificial Intelligence
But a way out of this dilemma was soon in sight: Parallel to introducing the dreaded DSGVO, the Berlin start-up ‚Brighter AI Technologies‘, published a unique solution to anonymise all visual data. “By using a deep-learning-algorithm we can remove relevant person-related data or even Personally Identifiable Information(PII) from videos and photos“, explains Marian Gläser, Co-Founder and CEO of Brighter AI, the principle of “Deep Natural Anonymization“. But it does not end with just removing data. Personally identifiable information such as faces or number plates is replaced by artificially replacing data generated by the system, and put into pictures and videos instead of the original PIIs. The artificial replacement data looks real; however, the critical PII is missing. In this way the data becomes compatible in accordance with the DSGVO. “Data once anonymised can be used by companies without having to ask the depicted persons for approval and for photo- and videos the right to be forgotten“, confirms Marian Gläser, “we enable companies to collect, save, and mutually use data in an unobjectionable manner.”
Brighter AI kills two birds with one stone: On the one hand the principles of data protection are observed by anonymising visual data. And on the other hand person-related data are available in their natural form and are thus not lost for future ‘Deep-Tech’ applications. On the contrary. They can be used for software development, marketing analyses or even be used to improve AI-algorithms. ”It is a great achievement for us to have found a solution, where data for modern analytics can be used like machine learning but which are still conforming“, says Gläser and hopes to contribute to develop artificial intelligence by using these anonymising approaches.
From Night to Daylight
Brighter AI, set up in 2017, as a spin-off of the car sub-supplier ‘Hella Inkubator’, uses the innovative deep-learning-method not only to anonymise data sets. Regulations on privacy and other legal requirements are by far not the only factors that influence visual data. Fog, rain, and other weather conditions disrupt the human vision and reaction just the same as bad lighting and darkness. They are responsible for difficult visibility and weather conditions which even the best camera cannot improve.
And this is, when the innovative deep-learning-method by Brighter AI - which asserted itself at the “NVIDIA’s GPU Technology Conference (GTC)“ against 1600 AI-innovators in October and was awarded to be ‘European’s hottest start-up‘ within the AI-sector - comes into its own: By using so-called ‘Perception Layers’, which remove visual restrictions for humans and machine, Marian Gläser and both co-founders, Chief Marketing Officer Asaf Birnhack and Chief Technology Officer Patrick Kern, have started to solve this basic problem of human vision. For this purpose, ‘Generative Adversarial Networks‘ (GANs) are used, whereby two neuronal networks, a generator and a discriminator work against each other and sort of try to surpass the results of the others. While for example the generator network turns a night photo taken with an infrared camera into a real daylight version, the discriminator network checks continuously whether this is a real or artificial photo. Up to seven neuronal networks are currently working in the system at the same time, learning from each other and optimising the results ongoing.
Secure thanks to Deep-Learning Applications
“Our system is already being used by the security industry”, says Gläser. It seems that this industry is predestined for this application as especially in the security sector sharp, unambiguous photos are required: darkness, fog, smog or bad lighting could massively affect the identification of criminals.
The Cloud-based API of the Brighter AI team, consisting of 12 people, transform data that can be understood much better and consequently, these sharper, clearer, in short - improved - photos will lead to make better decisions.
“At present we are concentrating on successfully carrying out various pilot projects in this area“, remarks Marian Gläser, who knows the great potential of his solutions, “as a perspective, this technology may also be applied to other situations, such as for example, in the automobile industry.” One of the ideas is, to build the Perception Layers into digital rear mirrors so that the rear vehicle area appears in daylight during the night. Functions that optimise camera-based driving assistant systems can be realised. Apart from the car industry, Brighter AI, whose turnover growth is estimated to be well over 70 million Dollars within the next five year, concentrates currently on finding solutions for the retail trade and the media industry.
Redefining limits for current and future cameras
Brighter AI is only right at the beginning, but the innovative Berlin company follows an important mission: “To redefine the restrictions of billions of existing and future cameras.” You do not need artificial intelligence or deep-learning-algorithms to understand that the future looks really bright.
Find more information about Brighter AI here: https://brighter.ai/