Each year, hundreds of person overboard (MOB) incidents occur across maritime sectors.
According to the UK Marine Accident Investigation Branch (MAIB), between 2015 and 2023, there were 308 reported person overboard (MOB) incidents — and tragically, 40% of them resulted in fatalities.
This alarming figure underscores the critical need for fast, reliable detection and immediate response to MOB events. Survival often depends on just a few minutes. Time is the most critical factor — survival rates drop dramatically after just 10 minutes in cold water.
SEA.AI’s MOB detection relies on a combination of thermal cameras, low-light RGB cameras, and custom-trained neural networks to detect and classify objects on the water surface — even in complete darkness or high-sea states.
The AI model is continuously trained on an extensive database of real maritime conditions, including images and videos collected worldwide, enriched with custom scenarios like simulated man overboard events.
SEA.AI’s system is designed to detect and classify:
✔ Floating objects like containers, buoys, and debris
✔ Vessels of all sizes, including non-AIS targets
✔ People in the water, whether moving or immobile (MOB detection)
✔ Marine life such as dolphins or whales
Unlike traditional MOB systems that rely on wearable devices or visual lookout by crew, SEA.AI performs continuous AI analysis on every pixel of the video feed — in real time, 24/7.
Whether it’s day or night, in calm waters or rough seas, the system maintains constant surveillance of the water surface to instantly detect a person overboard — and to continuously track their presence so they are not lost from view, even in challenging conditions.
This persistent tracking ensures crews can keep visual contact throughout the rescue process, significantly increasing the chances of a successful recovery.
✔ Human sight is limited in rough sea conditions.
✔ It’s nearly impossible to spot a person in the water at night
✔ The view from deck level is restricted compared to elevated systems like SEA.AI
✔ Lack of real-time tracking reduces search efficiency
SEA.AI overcomes these limitations with advanced AI-driven object recognition, ensuring constant monitoring of the water’s surface down to the pixel level.
To build a system capable of detecting people in the water with reliability, SEA.AI relies on real-world data acquisition, not just simulations or pre-existing footage.
That’s why the team regularly goes out to sea to collect raw visual data from actual man-overboard and near-collision scenarios.
Recently, during a testing campaign in the harbor of Brest in Brittany, the SEA.AI team conducted man-overboard simulations to collect raw visual data in real-life maritime conditions.
These trials focused on capturing people in the water under varying scenarios, both by day and night, to build a reliable dataset.
Once collected, this data is sent through a rigorous annotation process, where relevant objects and potential hazards — such as people, vessels, or floating debris — are manually labeled.
This annotated data is then used to train and refine SEA.AI’s neural networks, improving their ability to detect a person in the water with precision, even in visually complex situations.
Once the training is complete, the resulting improvements are rolled out through software updates, continuously enhancing the performance of SEA.AI systems already in operation.
While passive image collection is useful, it doesn’t always generate the optimal data needed for these high-stakes applications.
Since 2018, we have actively built custom, high-quality AI datasets tailored to the demands of maritime safety, setting new standards in marine machine vision.
By getting out on the water — even in cold and challenging conditions — the SEA.AI team is not just developing technology; they’re actively contributing to efforts to save lives at sea.
This hands-on approach ensures:
✔ Higher detection accuracy than standard AI models
✔ Optimization for unpredictable maritime conditions
✔ Ongoing improvements through real-world testing
A fast and reliable man overboard detection system can mean the difference between life and death.
With our AI-powered MOB detection, search and rescue teams gain:
As maritime safety evolves, AI-driven MOB detection is no longer optional—it’s essential.
We at SEA.AI remain at the forefront, pioneering solutions to ensure no distress signal goes unnoticed.
We are proud to announce that our Watchkeeper system has been awarded the Innovation Route Award at the Cannes Yachting Festival in the category “Solutions that improve the navigation experience.”
A new office has been opened on the Gold Coast to support growing interest in AI-powered marine vision systems across the Asia-Pacific region.
Affordable, accessible AI-driven safety and security takes center stage in Cannes.
