At least since ChatGPT, artificial intelligence has become indispensable in many professions and areas of life. At the same time, you can often find them in places where you would not have expected them at first – in the swimming pool, for example. TECHBOOK reveals how AI can save swimmers’ lives in the future.
The Kleinfeldchen outdoor and indoor swimming pool near Wiesbaden has had to contend with the same adversities as many other swimming pools in Germany in recent years: first the pandemic, which kept visitors away, then the energy crisis and finally the shortage of skilled workers. An estimated 3,000 professionals are missing in German pools, many of whom are trained lifeguards who supervise bathers and save lives in an emergency. However, if they cannot adequately meet this great responsibility due to the lack of staff, the pools have to reduce their opening hours. Because even under supervision, tragic accidents happen again and again. So gave the DLRG known that a total of 13 people drowned in swimming pools in 2022. In the Kleinfeldchen swimming pool, of all things, the use of AI should prevent such drowning accidents in the future.
In an emergency, a few seconds count
Especially on hot days and in the summer holidays, numerous visitors flock to the outdoor and indoor pools. But even the most vigilant lifeguard can easily miss a potentially dangerous situation. The Bad Kleinfeldchen can accommodate 4000 to 6000 guests at the same time on a full day. That’s where it can get confusing. If the lifeguards also look at the water at an angle during their patrols at the edge of the pool, the surface often reflects the light so strongly that they can hardly see the bottom. In the case of a water rescue, however, just a few seconds can be decisive. After three to five minutes without oxygen, an adult’s brain suffers irreversible damage. Lifeguards must therefore recognize an emergency very quickly in order to be able to act in good time.
In order to ensure the safety of bathers even better, Mattiaqua, the municipal operator of the Kleinfeldchen pool, has come up with an unusual solution: an AI is supposed to make the pool safer. Since 2020, a so-called drowning detection system of the start-up has been monitoring Lynxight one of the pools. The name Lynxight is a play on the English words lynx sight, to German lynx eye. To implement the system, four surveillance cameras were mounted on the ceiling above the 25 by 15 meter swimming pool. “The cameras detect the movements in the water and record a movement profile that is analyzed by the AI,” explains Thomas Baumthe manager of the Mattiaqua Baths.
Also interesting: The 11 most popular outdoor pools in Germany
How the “lynx eye” works
To protect visitors’ privacy, the AI only recognizes the outlines of the swimmers and focuses on their movements. If the AI detects unusual movement patterns, it sends a sound and vibration alarm to the swimming champions’ smartwatch. An image of the pool and a red dot on the watch’s display indicate exactly where the potential emergency is. In addition, the AI presents three real-time photos to help the lifeguards assess the situation.
After the first test phase, this additional monitoring instance proved to be a helpful addition for the lifeguards. However, some teething problems had to be overcome first. At the beginning, the system indicated typical but irregular swimming movements, such as the roll turn at the end of a lane, as an emergency. But this is where the strengths of the AI come into play: After each alarm, the lifeguards can give feedback with a simple thumbs-up or down-click as to whether the system’s assessment was correct or incorrect. In this way, the AI can constantly improve and save maneuvers such as the roll turn as a harmless movement pattern. In addition to the practical training in Bad Kleinfeldchen, the AI also learns on the basis of existing data from other countries. After all, the movement patterns that typically precede a drowning accident are the same worldwide.
Sudden exhaustion, dizziness, unconsciousness or collapsing can also affect trained swimmers, so the use of AI systems in deep sports pools makes sense. The AI is therefore particularly trained to recognize the characteristic up and down movements when a swimmer suddenly has problems keeping his head above water. The AI also alerts if it is underwater for a long time and if it detects that a toddler is moving away from an adult.
Use of AI is to be expanded
Since the test phase in the Kleinfeldchen outdoor and indoor pools ended with very convincing results, operations manager Thomas Baum now wants to equip the outdoor pool and the non-swimmer area with AI systems as well. Especially in confusing areas and vulnerable groups, i.e. non-swimmers and children, the AI should help to identify possible dangerous situations at an early stage.
Bad Kleinfeldchen and the start-up from Israel
The drowning detection system was developed by the Israeli start-up Lynxight. In 2019, the two founders Eyal Amit and Omer Bar-Ilan developed software that analyzes human movements in water. Since water, as a moving and reflective medium, is relatively difficult for an AI to analyze, the algorithm must first calculate these disruptive factors. In order to better classify human movements, the AI is also constantly learning on the basis of anonymous video data from other swimming pools.
While there have been other forms of automatic drowning detection, these have been less practical. One approach relies on underwater cameras. However, their installation and maintenance is expensive and time-consuming, especially when the cameras have to be retrofitted in existing swimming pools. Another concept works with sensors in bracelets that all visitors should wear, similar to the locker key. The sensor sounds an alarm if it is under water for too long. Here, too, the cost of materials and application are rather unpractical. Lynxight, on the other hand, relies on three simple components: ordinary cameras, smartwatches and WLAN. All elements can be easily retrofitted in the swimming pools without anyone from the Israeli team having to be on site. In this way, the Lynxight system is now being used in seven countries.
With all the advantages of AI monitoring, it could become the new safety standard in German swimming pools in the future – at least that’s what it is Florian Grojer from Lynxight convinced. He compares the current situation with aviation fifty years ago. After all, at that time one could not have imagined that the autopilot would one day be part of the standard equipment, and today it is impossible to imagine life without it.
Also read: The 6 best sports apps for more exercise in everyday life
AI in the swimming pool as the new normal?
In times of a shortage of skilled workers, AI monitoring is a helpful relief for the lifeguards – but one that also has a significant financial impact. The AI system costs between 30,000 and 40,000 euros – per pool. The costs vary depending on the pool size and number of cameras. This means that the pool operator pays just as much for the AI assistant as for a trained full-time employee with a collective agreement. For Thomas Baum, however, the calculation is still worthwhile: “If it only works once in ten years and saves a human life, then every cent invested was worth it.” 14-year-old boy died while swimming at school. Unlike the pool in the public swimming pool, the training pool was not equipped with the AI monitoring system.
The reason why such AI systems are not yet more widespread is, on the one hand, the financing problems that many pools already have. On the other hand, some pool operators still have doubts about the safety standards, which they consider to be insufficiently mature. Nevertheless, the Kleinfeldchen outdoor and indoor pool also serves as a model for other pools. Stadtwerke München, in cooperation with Lynxight, has also started a two-year “Smart Swimming Pool” test phase in Südbad. The first experiences are very positive. However, the AI still has to learn to distinguish potential emergencies from the “dead man”. In this swimming exercise, swimmers float on the water surface on their backs and the maneuver is particularly popular with sunbathers. An AI test phase is also underway in the CabrioLi leisure pool in Lippstadt, and the Nordbad in Darmstadt is also planning to use AI systems.
Also interesting: 12 thermal baths and baths where nudists can be among themselves
And what about data protection?
The transparency with which the bathrooms explain and point out the AI surveillance varies greatly. The same applies to handling the video material. While the AI in Munich processes the images in real time and does not take any pictures of the visitors, the operator of Kleinfeldchen stores the pictures for a certain period of time. In the event of an emergency, you want to be able to analyze the recordings afterwards. Apart from that, Thomas Baum assures that the data is stored securely and in compliance with the GDPR.
In view of the positive experiences, it could well be that AI systems will support the lifeguards in more and more swimming pools in the future. However, the jobs of the lifeguards are not threatened by the drowning detection systems. After all, the AI can neither jump into the water nor rescue swimmers who have had an accident.