Researchers from the Norwegian Veterinary Institute and the Norwegian Institute for Nature Research have developed a deep-learning model that can distinguish wild from farmed Atlantic salmon with 95% accuracy. By analyzing nearly 90,000 scale images, the AI system addresses a critical conservation challenge in Norway, where an estimated 300,000 farmed salmon escape annually, threatening wild populations that have declined by over 50% since the 1980s.
Imagine trying to spot the difference between two nearly identical fish—one born to navigate Norway’s wild rivers and another raised in the controlled environment of a fish farm. For conservation biologists, this isn’t a trivial puzzle but a pressing ecological crisis. Now, artificial intelligence is providing a solution that could transform how we protect vulnerable species.
Scientists have unveiled a breakthrough deep-learning tool that can identify whether a salmon is wild or farmed simply by examining images of its scales. The research, published in Biology Methods and Protocols, comes at a critical time for Norway’s wild salmon populations, which face unprecedented threats from escaped farmed fish.
Why does this matter? Norway produces over 1.5 million metric tons of farmed Atlantic salmon annually, but each year approximately 300,000 of these fish escape into the wild, reported Biology Methods and Protocols. These escapees don’t just swim away—they compete with wild salmon for food and spawning grounds, introduce diseases like sea lice, and interbreed with native populations, weakening their genetic fitness.
READ ALSO: https://www.modernmechanics24.com/post/ai-agent-that-turns-sketches-into-3d-cad-designs
“The decline of wild Atlantic salmon in Norway represents one of the most significant conservation challenges in modern aquaculture,” the researchers noted in their paper. What makes this particularly troubling is that genetic analysis shows approximately two-thirds of wild salmon in Norway now carry genetic signatures indicating interbreeding with farmed varieties.
Until now, monitoring this problem required painstaking manual examination of fish scales. Like tree rings, salmon scales form concentric circles that record the fish’s growth history. Farmed salmon show regularly spaced rings reflecting consistent feeding and controlled conditions, while wild salmon display irregular patterns mirroring seasonal variations in temperature, prey availability, and migration.
According to Biology Methods and Protocols, the research team established a standardized processing pipeline and trained a convolutional neural network using scale images from hundreds of Norwegian rivers dating back to the early 1930s. The massive dataset included approximately 90,000 images, with farmed salmon comprising about 8.5% of the total.
WATCH ALSO: https://www.modernmechanics24.com/post/us-marines-seaglider-rescue-showcase
The results were striking. The AI model achieved 95% accuracy in distinguishing farmed from wild salmon across most Norwegian rivers from 2009 to 2023. Perhaps most importantly, the system provides confidence estimates with each prediction, allowing researchers to understand how certain the AI is about its classification.
This technological advancement couldn’t be more timely. Atlantic salmon abundance in Norway has declined by more than 50% since the 1980s and now sits at historically low levels. The escaped farmed salmon problem compounds other threats wild populations face, including climate change and habitat degradation.
The deep-learning tool offers more than just identification—it provides a scalable solution to a problem that previously required enormous human resources. Manual scale examination is not only time-consuming but extremely expensive, limiting how comprehensively scientists can monitor salmon populations across Norway’s vast waterways.
READ ALSO: https://www.modernmechanics24.com/post/china-unveils-as700-airship-in-shaoxing
What does this mean for conservation efforts? The AI system enables rapid processing of thousands of scale images, giving researchers near real-time data on the extent of farmed salmon infiltration into wild populations. This information is crucial for developing targeted strategies to protect vulnerable wild salmon stocks before they’re irreversibly compromised.
The research represents a perfect marriage of ecological science and cutting-edge technology. By leveraging historical data going back nearly a century while applying modern deep-learning techniques, scientists have created a tool that respects the past while embracing the future of conservation.
As aquaculture continues to expand globally, the challenges facing wild fish populations will only intensify. The Norwegian team’s work demonstrates how artificial intelligence can help balance human food production needs with the urgent requirement to protect biodiversity. Their salmon identification system may soon become a model for conservation efforts worldwide, proving that sometimes the most powerful tools for protecting nature come from the most unexpected places.
WATCH ALSO: https://www.modernmechanics24.com/post/us-supersonic-jet-cuts-flight-time-silences-sonic-boom
