From inside a package about the size of a shoebox mounted to a ship’s deck, a set of highly stabilized heat-sensing cameras scans the ocean’s surface. Suddenly, against the misty waves far in the distance, they spot a small puff of white. And another. Now the algorithm catches on. A machine learning system snags the footage and runs it through a neural network trained on millions of similar snippets.  

Comparing what it’s detecting against its training data, the artificial intelligence model makes a call: that small burst of heat in the distance is a spout of whale breath. The computer system pings a remote expert on standby who double-checks the machine’s work. Within a minute, the expert forwards the alert back to the ship, catching the captain’s attention with enough time for the crew to change course and, hopefully, avoid the whale becoming maritime roadkill.

This is WhaleSpotter, an artificial intelligence-powered whale detection system that aims to transmit real-time alerts to ships to prevent them from colliding with whales—a threat that leads to the injury or death of thousands of whales each year.

Video courtesy of WhaleSpotter

Led by Daniel Zitterbart, a biophysicist at Woods Hole Oceanographic Institution in Massachusetts, scientists have been testing this new AI-powered but human-verified whale detection system on ferries, research vessels, and cruise ships, and from land-based installations along the east and west coasts of North America, as well as in parts of the Southern Ocean.  

Since WhaleSpotter first got underway during research trials in 2019, its capabilities have grown tremendously. Across its more than two dozen ship- and land-based installations, the global WhaleSpotter network made more than 51,000 marine mammal detections in 2024, up from just 78 its first year. All of those detections were automatically sent to a remote data center in real time, but only a few ships have opted in to receiving the 24/7 alerts.

But those ships are modestly sized. And in the quest to save whales from deadly collisions, one of the greatest challenges is protecting them from some of the biggest—and most common—vessels at sea: container ships.  

Strikes from container ships, which are massive and hard to maneuver, are one of the leading causes of death for large whales, according to Zitterbart. Peering out from a cargo ship’s bridge high above the waves, especially at night or in fog, a captain may struggle to see a whale soon enough to shift the course of the vessel, which can easily be 250 meters (800 feet) long. That’s why Zitterbart’s team recently began a research partnership with the Hawai‘i-based Matson Navigation Company to start adapting WhaleSpotter’s technology for this key class of vessels. 

Tailoring WhaleSpotter to work on container ships has required special considerations. Slower and harder to turn, container ships need more advance notice than other craft. However, container ships also tower over the ocean. Making use of the higher vantage point, Zitterbart and his team have been able to increase the distance at which their system can reliably spot whales. Testing longer-range cameras and adjusting the stabilization system on Matson’s container ships plying routes along Hawai‘i, Alaska, and the U.S. west coast, the team found that the technology can now reliably spot marine mammals up to 6 kilometers (nearly 4 miles) away. Matson’s ships are not yet receiving real-time alerts through WhaleSpotter’s systems, but Zitterbart and his colleagues continue fine-tuning their detection system to get it ready for prime time. 

“I think we’re almost there,” says Zitterbart.  

“We are excited by the early results,” adds Keoni Wagner, a Matson spokesperson. “Assuming the system achieves current expectations, we plan to expand use to our entire domestic fleet.”

From the perspective of John Calambokidis, a marine biologist with the nonprofit Cascadia Research Collective, there are, broadly, three strategies for reducing incidences of ships hitting whales: shift vessels’ routes, slow them down, and use real-time detection to avoid whales. Calambokidis says the third strategy has not received nearly enough attention, and that Zitterbart’s approach provides an important contribution. “That’s no simple feat,” he says of the expert-reviewed AI approach. He adds that the reliance on thermal cameras—which detect heat rather than light—makes the system particularly useful at night, when many whale species are more likely to be near the surface than during the day.  

WhaleSpotter, which has just spun off into a for-profit company, isn’t the only AI-enhanced thermal camera system able to detect whales. Awarion and SEA.AI can, too. But Zitterbart contends that what sets his technology apart is that WhaleSpotter is purpose-built for marine conservation. As such, he’s adamant about having humans validate the machine’s work. “Many people said, ‘Isn’t that overkill? Can’t we get rid of that?’” Zitterbart says.

While the AI system is designed to filter out false alarms—such as signals from birds, breaking waves, and boats—Zitterbart’s aim is for ship captains to receive zero false alerts, so that every ping truly requires their attention. Removing human oversight risks flooding ship captains with false reports, which could lead to frustration and alert fatigue. At risk is the very survival of species like the North Atlantic right whale, an endangered animal that has suffered heavily from ship strikes and has only 370 individuals left: “We cannot afford to ever miss an animal,” he says.

Calambokidis emphasizes that preventing collisions between whales and ships requires using multiple, complementary strategies. While Zitterbart readily agrees that WhaleSpotter is no silver bullet, he says it’s particularly suited to certain goals—like limiting the deaths of endangered species.

Ultimately, he wants more ships to carry WhaleSpotter cameras. “The true power will come to life once there are hundreds of vessels using this tech,” he says. “Then the collected information can be shared in real time with vessels not using the technology, too.”

Yet as he works to grow WhaleSpotter’s reach, Zitterbart’s focus remains on the animals: “Every single whale that is not struck because of the technology is a success.”