We are all working on ways to comply with the latest corona guidelines. Whether it is on a terrace, at the office or in schools; life is getting going again and we have to find a new and safe way to move forward. Not surprisingly, corona apps spring up like mushrooms. There is only one recurring theme: What about privacy? However, privacy can be guaranteed with the help of AI (artificial intelligence).
If you want to measure whether you have been around someone with corona, you must have an app on your phone. But even if you want to reserve a place at the office, you can do that with your phone. Gone privacy. But what if you turn it around and analyse the risk of a space rather than a person’s risk? That is exactly what SmartEagle does with the Distance Sensor. The people do not measure their distance with their own devices. Sensors in this space measure the distance between people, whoever they may be.
AI teaches a sensor what a person is and what a plant
The SmartEagle Distance Sensor uses optical sensors. The data from the different sensors are combined with each other to get an overall picture of the occupation of the space. Artificial Intelligence (AI) has been used to do this. But how does a sensor know what a person is and what a plant or a chair?
Machine Learning models
The existing software of SmartEagle (to find workplaces and monitor occupancy in an office) has already been trained by Machine Learning (ML) to recognize people, coffee cups, desks, laptops and plants, among others. For the Distance Sensor it is particularly important that the software knows the difference between people and everything else. The model has now been trained in this. On location, the model will be further adjusted to the specific environment, for example taking the dimensions of the space into account to determine the distances between people. SmartEagle makes as much use as possible of existing (public) data files and existing ML models, because this improves the transparency and maintainability of the software.
How does that work?
The optical sensors register images, but do not store them in a database. The visual data is directly coded thanks to ML (read: It is a human, laptop, telephone or something else) and then only this data is put in the database. Identification of a specific person is not possible based on these data. The distance between people is linked to this, creating a real-time insight into the situation. Administrators can set which distances are permissible in which situations. Six feet is now the norm, but this differs per country and companies are of course free to choose a greater distance. Also in places where there is a lot of movement, a preventive choice can be made for a greater distance.
Contagiousness of a room
In this way, based on the statistical distribution of the distances, an estimate can be made of the “contagiousness” of the space. With a relatively large number of shorter distances between people over a longer period of time, the risk of contamination is greater than with a one-off short distance between two people. How the outcome of the software is then communicated to the users of the space itself or the person responsible for the space, can be completely set up according to your own wishes. For example, a message could be sent to the smartphones of those present (who have registered), a screen can indicate how safe the situation is or a doorman can refuse people at a certain level.
An important advantage of the SmartEagle Distance Sensor is the privacy of users. The software does not identify people, it does not use facial recognition and it does not us someone’s smartphone. The sensors purely determine the number of people in a room and their mutual distance, and with that the chance of contamination in that room.
Recently, research by three universities and RIVM revealed that privacy is a major issue for Dutch people. The studies showed that the willingness of the Dutch to use a corona app with tracking and tracing has declined in recent weeks. By using Machine Learning and Artificial Intelligence, it is possible to develop apps that protect us, but do not violate our privacy. Perhaps not as a broadly applicable app for the entire Dutch society, but within smaller “societies” such as companies, care and educational institutions.