Thanks SRF! AVA-X in Rundschau

The extent of some technological upheavals in recent years has caused fear among many people. We are very pleased that Swiss Radio & Television has addressed the topic of automatic facial recognition.

English edition — originally published in German as Danke SRF! AVA-X in der Rundschau.

Introduction

The extent of some technological upheavals in recent years has caused many people to fear that their freedom, security, and autonomy will be restricted. Indeed, many modern inventions are associated with uncertainties, and we are very pleased that Swiss Radio & Television has also addressed one of these topics: automatic facial recognition.

SRF Coverage

SRF reported on several channels (radio, article, podcast) about the football club Brøndby, which uses cameras and facial recognition software to identify fans with stadium bans and deny them access. The unsuccessful use of facial recognition at the 2017 Champions League Final in Cardiff was also mentioned, where the so-called False Positives (authorized entrants mistakenly classified as unauthorized by the system) were far too high. The decisive influence of video recording quality is highlighted: if recording conditions are better (lighting, view of face, etc.), the error rate decreases. According to the Federal Data Protection and Information Commissioner (FDPIC), facial recognition software is generally permissible in sports stadiums.

SRF simultaneously emphasizes that the various stakeholders still see open legal questions. Furthermore, some participants also express doubts as to whether automatic facial recognition is worthwhile given the observed error rates.

SRF's reporting points out significant opportunities and challenges in the use of facial recognition in sports stadiums.

Positive Trend

The example of the Champions League Final only tells the beginning of the story. In South Wales, the experiment was not abandoned but pursued with persistence, as official figures and the associated theme page show.

What is new in terms of data protection law?

It is important to distinguish between facial recognition and data recording. Data is already being collected on a large scale today within the existing legal framework. This applies, for example, to ordinary video surveillance or the recording of personal data of disruptive fans. However, data protection issues related to automated facial recognition concern other data: it is about establishing relationships between the data.

Prospects for Switzerland

From an economic perspective, artificial intelligence offers enormous potential worldwide. If Switzerland hesitates in its development, it will destroy the basis for technical expertise, innovation, and ultimately industry in Switzerland.

Fairer, Cheaper, Faster

It must be clarified that authorized visitors who have been wrongly classified (False Positives) are not simply banned from the stadium. Access control is an overall process in which a warning by automated facial recognition merely leads to the usual, manual personal check being carried out. The introduction of facial recognition therefore leads in particular to an increase in the efficiency of the stadium operator, an increase in security, and less waiting time in queues.

Deep Impact and Facial Recognition

At Deep Impact, we have been researching facial recognition for years. In internal measurements, our algorithms regularly confirm their position at the top of the world. For example, we measure an accuracy of 99.82% in the well-known "Labelled Faces in the Wild" (Unrestricted, Labeled Outside Data) benchmark. We continuously optimize the software and can also scale this accuracy excellently.

AVA X and Sentinel under Stress Test in Turkey

We tested our automated facial recognition at high-risk football matches in the Turkish National League. We achieved an optimal result: not a single False Positive.

We know the topic of artificial intelligence from the ground up. Artificial intelligence is not a horror scenario. It is exclusively a collection of mathematical tools that offers a practically implementable and immense added value for countless applications in business and private life.

Conclusion

Deep Impact supports companies in their digital transformation with tailor-made solutions and proven expertise.