Lifeguards have their work cut out for them keeping an eye on a crowded beach. Staying attentive for hours on end and quickly recognizing an anomaly in a sea (no pun intended) of noise is not a task humans are particularly well suited to. If only there was a machine that was very good at detecting patterns, that never got tired, and was never distracted by a skimpy swimsuit.
Well, of course there is, and a start-up called Sightbit has developed an AI-driven computational approach to assist lifeguards in detecting distressed swimmers. They operate under a monthly software-as-a-service subscription model and handle the initial system setup onsite. Sightbit is currently being pilot tested at a beach known for dangerous rip currents that lies south of Tel Aviv, Israel.
Several cameras are positioned around the beach, and the imaging data is fed into an NVIDIA Jetson AGX Xavier located in the lifeguard towers. The image data is classified by a convolutional neural network that was trained on tens of thousands of images. When the system detects a swimmer in trouble it highlights the swimmer, and adds a text bubble displaying the situation detected, on a screen that lifeguards can monitor. Clicking on a text bubble zooms in to give a clearer look at the situation.
In addition to detecting swimmers in distress, Sightbit can also detect unattended children and ocean hazards (e.g. rip currents). Once a situation is detected, it is then up to the lifeguards to do the real work of rescuing swimmers.
However, in a situation where lifeguards cannot be present, Sightbit may someday offer some help. For example, if a swimmer were caught in a rip current, it may be possible in the future to extend the system to dispatch a drone that could then drop the swimmer a flotation device until further help arrived.