Fish Facial Recognition: Technology to Help Fish Stocks

The most accessible of these systems is the recreational angler app, Fishbrain. By borrowing Google’s image recognition technology, Fishbrain allows users to upload photos of their catches and immediately gives a 90 percent accurate analysis of the species they’re holding. I tested this app feature on several of my catches and was impressed with its accuracy. 

Simply put, Google uses a mathematical model to determine shapes, colors, lines and proportions to scan an index and find matching results. This has been a game changer for anglers who are uncertain of which species they’ve caught. App subscribers aside, Fishbrain has teamed up with the US Fish and Wildlife Service, the Florida Fish and Wildlife Conservation Commission and several other organizations to provide real-time data when endangered or invasive species are captured. The ability to document such species could potentially help scientists determine if action needs to be taken in specific waterways.

The Wild Salmon Center (WSC) based in Portland, Oregon is taking fish recognition one step further. Their focal point is on taimen in the Russian Far East, specifically the Sakhalin and Siberian taimen. They are the largest salmonids in the world and can live up to 30 years and weigh over 100 pounds. Despite active conservation measures and tagging efforts, taimen in Russia and Mongolia continue to decline throughout their range, with some stocks plummeting as much as 70 percent from their historical baseline. Organizations like the Wild Salmon Center, the Taimen Fund and the University of Nevada are committed to learning more about taimen life history — and they’re using facial recognition software to do it.

With the help of funding from National Geographic Society, the Wild Salmon Center’s director of science, Dr. Matt Sloat, is working closely with guides from Konin Lodge on Russia’s Tugur River to accumulate data about individual fish. Guides photograph and measure the fish their clients catch and document the location before passing the information on to Sloat. From there, Sloat enters the information into a database, where it is stored in case the fish is caught again. For the WSC, the photo identification work is the first part of quantifying the value of protected areas for taimen and other salmonids so that they can support their Russian partner’s efforts to create two new protected areas by the end of 2019. 

“We use these tools to estimate the abundance of taimen within and outside of protected areas, which is a way to quantify the benefit of habitat protection for taimen,” Sloat said. “This bolsters support for efforts to create additional protected areas, which benefits taimen as well as rare species like the Amur tiger that share some of the same watersheds.”

The computer program they’re using to do this work is called WildID, pattern recognition software designed to store and identify each fish’s unique spotting configuration. The system scans the fish from it’s snout to it’s pectoral fins, determining it’s spot pattern and keeping a record in the system. The hope is that documenting individual fish will help to determine the Tugur River’s taimen population, movement patterns, size ranges and differences between males and females.

The project is still in its infancy (the WSC completed the program’s pilot season in June 2017), but the approach is indicative of a new age. Programs around the world have historically relied on the cooperation of guides to help with tagging efforts, but this can require training and is only useful when guides are willing to participate. While not all guides are eager to add tagging tasks to their daily duties, they’re used to taking pictures of client’s fish, so it’s an easy transition. It’s a win-win.

Developed at Dartmouth College, WildID is one of several facial recognition software programs used by scientists to document the distinguishing features of various animals. The concept isn’t new; for years, whales and other mammals have been identified by unique markings and physical characteristics. Programs like WildID now offer an alternative to additional handling while tagging fish and even practices like tranquilizing mammals. The software has become an essential tool for field biologists who use camera traps to study and monitor terrestrial animals and birds.  

Though WildID is primarily used by scientists, the software is free and available to everyone. Dr. Sloat mentioned that there is a small amount of technical work necessary to convert collected data into usable information and that photos should be standardized to avoid confusion, but that the program should work on all salmonids and fish species with unchanging spotting patterns. In his experience, the software has been 99 percent accurate.

The most recent application of fish facial recognition is its role in high-tech salmon farms. BioSort vision recognition system helps diagnose fish infected with sea lice. Once corralled, fish are guided to a camera that uses the same concept as WildID. From there, the technology goes one step further by picking up on abnormalities like lice and ulcers. If the fish is assessed as infected, it is guided into a holding pen where it will receive individual treatment.

It is standard practice in aquaculture to treat the whole farm if even one fish is hosting parasites, but this new technology allows individual fish to be treated instead. Sea lice passing from farmed to wild fish is a highly controversial element of the fish farming business, but this new technology could go a long way to preventing that transmission and many people are waiting to see how well it works.

While it would be nice to live in a world where fish stocks are so abundant we don’t need to monitor them, the reality is that many need all the help they can get. If we’re going to continue to catch and release fish for our own pleasure, it only makes sense to use the tools available to us to collect useful information. Facial recognition in fishing is still too new to make a major impact in most of our lives but with the rate technology is advancing, it wouldn’t surprise me if anglers will soon be using facial recognition apps to log data. We can still run away to the river to escape the modern world, but technology is following closely behind us.

Feature Image by Dave McCoy