Artificial Intelligence As Tool for Animal Spotting: How Does AI Recognize Individuals From Different Species of Whales?

With ecosystems in crisis, engineers and conservationists team up to enable the public to identify animals using crowdsourced data. In a breakthrough, artificial intelligence can help people recognize the animal they are looking at, aiding in environmental conservation.

Challenges in Photo Identification of Species

Many aspects of animal ecology can be investigated using noninvasive photo identification. This strategy is becoming more efficient since matching individuals between images is increasingly automated.

However, the convolutional neural network models that facilitate this change require many training images to generalize well. Because of this, they have often been designed for individual species that satisfy this threshold. Such single-species methods could underperform since they ignore potential similarities in recognizing characteristics and the photo-identification procedure among species.

Animal Recognition With WhaleID

It is estimated that 13 million people go whale watching worldwide each year. The industry is said to be worth $2.1 billion, yet the chances of spotting whales are very low. On most whale-watching trips, observers can only see 2% of the whale for 2% of the time. According to whale conservationist Ted Cheeseman, huge animals do not patiently pose for photos. Nevertheless, going out in a boat to spot a whale remains very popular.

The humpback whale is the main attraction of the whale-watching business since it is relatively common and spends time on the water's surface. It is also very engaging, allowing watchers to catch a whale's breach or a flick of its tail.

To help whale watchers feel a closer connection to each animal, Mr. Cheeseman's research company, HappyWhale, developed an image recognition algorithm called WhaleID, which enables users to upload their photos onto its website. Its artificial intelligence software then quickly sifts through its database containing over 70,000 different whales to identify the exact animal spotted by the user.

The algorithm can tell the whale's name if it has already been given one or if the user is invited to choose one. If the observed whale has been previously recorded, the algorithm will show a map of all locations where it has been sighted. The artificial intelligence uses adapted human facial recognition software to identify the whale's tail that has been uniquely shaped, colored, or marked.

According to Cheeseman, the AI is so effective that it does not need a whale's tail to remain still and spread out. It can identify the animal even if the tail is curved, rotated, or in the act of diving.

The WhaleID system is free for the public, but a monthly subscription is required from whale-watching operators and cruise ships. The company noted that they charge for commercial use of the mobile app because guides are expected to instantly identify a whale and tell its story for whale watch companies and expedition cruise operators. HappyWhale believes this adds value and generates enthusiasm, allowing the guide to tell a true story about the whales they are encountering in a way that connects deeply with whale watchers.

The company also uses all the information from uploaded whale photos to help track its numbers and movements. This technology is an example of a growing trend where conservation groups use artificial intelligence to allow the public to identify animals.