A team in Israel is making a splash with AI.
It started as biz school buddies Netanel Eliav and Adam Bismut were looking to solve a problem to change the world. The problem found them: Bismut visited the Dead Sea after a drowning and noticed a lack of tech for lifeguards, who scanned the area with age-old binoculars.
The two aspiring entrepreneurs — recent MBA graduates of Ben-Gurion University, in the country’s south — decided this was their problem to solve with AI.
“I have two little girls, and as a father, I know the feeling that parents have when their children are near the water,” said Eliav, the company’s CEO.
They founded Sightbit in 2018 with BGU classmates Gadi Kovler and Minna Shezaf to help lifeguards see dangerous conditions and prevent drownings.
The startup is seed funded by Cactus Capital, the venture arm of their alma mater.
Sightbit is now in pilot testing at Palmachim Beach, a popular escape for sunbathers and surfers in the Palmachim Kibbutz area along the Mediterranean Sea, south of Tel Aviv. The sand dune-lined destination, with its inviting, warm aquamarine waters, gets packed with thousands of daily summer visitors.
But it’s also a place known for deadly rip currents.
Sightbit has developed image detection to help spot dangers to aid lifeguards in their work. In collaboration with the Israel Nature and Parks Authority, the Beersheba-based startup has installed three cameras that feed data into a single NVIDIA Jetson AGX at the lifeguard towers at Palmachim beach. NVIDIA Metropolis is deployed for video analytics.
The system of danger detectors enables lifeguards to keep tabs on a computer monitor that flags potential safety concerns while they scan the beach.
Sightbit has developed models based on convolutional neural networks and image detection to provide lifeguards views of potential dangers. Kovler, the company’s CTO, has trained the company’s danger detectors on tens of thousands of images, processed with NVIDIA GPUs in the cloud.
Training on the images wasn’t easy with sun glare on the ocean, weather conditions, crowds of people, and people partially submerged in the ocean, said Shezaf, the company’s CMO.
But Sightbit’s deep learning and proprietary algorithms have enabled it to identify children alone as well as clusters of people. This allows its system to flag children who have strayed from the pack.
Rip Current Recognition
The system also harnesses optical flow algorithms to detect dangerous rip currents in the ocean for helping lifeguards keep people out of those zones. These algorithms make it possible to identify the speed of every object in an image, using partial differential equations to calculate acceleration vectors of every voxel in the image.
Lifeguards can get updates on ocean conditions so when they start work they have a sense of hazards present that day.
“We spoke with many lifeguards. The lifeguard is trying to avoid the next accident. Many people go too deep and get caught in the rip currents,” said Eliav.
Cameras at lifeguard towers processed on the single compact supercomputing Jetson Xavier and accessing Metropolis can offer split-second inference for alerts, tracking, statistics and risk analysis in real time.
The Israel Nature and Parks Authority is planning to have a structure built on the beach to house more cameras for automated safety, according to Sightbit.
Palmachim Beach lifeguards have a lot to watch, especially now as people get out of their homes for fresh air after the region begins reopening from COVID-19-related closures.
As part of Sightbit’s beach safety developments, the company had been training its network to spot how far apart people were to help gauge child safety.
This work also directly applies to monitoring social distancing and has attracted the attention of potential customers seeking ways to slow the spread of COVID-19. The Sightbit platform can provide them crowding alerts when a public area is overcrowded and proximity alerts for when individuals are too close to each other, said Shezaf.
The startup has put in extra hours to work with those interested in its tech to help monitor areas for ways to reduce the spread of the pathogen.
“If you want to change the world, you need to do something that is going to affect people immediately without any focus on profit,” said Eliav.
Sightbit is a member of NVIDIA Inception, a virtual accelerator program that helps startups in AI and data science get to market faster.
The post Sand Safety: Startup’s Lifeguard AI Hits the Beach to Save Lives appeared first on The Official NVIDIA Blog.