AI Surpasses Biologists in Identifying Salmon Lice with 97.5% Accuracy

On July 4, 2026, researchers from the Norwegian University of Science and Technology and Wageningen University revealed that trained AI models now outperform experienced biologists in spotting salmon lice. This breakthrough was published in the journal Computers and Electronics in Agriculture, highlighting the urgent need to combat this persistent threat to both farmed and wild salmon populations.

AI's Efficiency in Salmon Lice Detection

Researchers analyzed over 120,000 images of salmon lice larvae, training AI models that demonstrated remarkable speed and precision. The study found that while biologists spent more than 30 hours to identify 82% of the larvae in a complex seawater sample, the AI accomplished the same task in just 30 minutes, achieving an accuracy rate of 97.5%.

According to researcher Lars Christian Gansel, head of NTNU's Department of Biological Sciences, this new method could revolutionize how the salmon industry manages lice infestations. “More information is needed about the spread of salmon louse larvae to document the effectiveness of current methods,” Gansel stated.

The Salmon Lice Crisis in Aquaculture

The salmon lice, specifically Lepeophtheirus salmonis, have thrived due to the booming fish farming industry, which has increased the availability of host fish. This tiny parasite poses a significant risk to wild salmon stocks, making it one of the industry's most pressing challenges. Fish farms release between 400 million and 450 million salmon and rainbow trout fry annually into Norwegian waters, leading to millions of lice larvae being dispersed daily.

The industry is governed by a traffic light system aimed at protecting wild fish. Facilities receive ratings based on the number of lice present, which dictates whether they can increase, maintain, or decrease production.

Challenges in Monitoring Salmon Lice

Counting salmon lice larvae in the ocean presents significant difficulties due to the vast quantities of other organisms present. Gansel emphasized that salmon lice are relatively rare compared to other plankton, necessitating the analysis of large water volumes to obtain accurate counts.

Historically, methods for monitoring salmon lice have been cumbersome and costly. “Currently available camera systems for plankton analysis often lack the resolution needed to distinguish between species and developmental stages,” Gansel noted. The integration of artificial intelligence and machine learning into this field offers a promising solution.

• AI model identified 97.5% of larvae in 30 minutes.
• Biologists took over 30 hours to identify 82% of larvae.
• Over 120,000 images used to train AI.
• Fish farms release 400 million to 450 million fry annually.