HyperAIHyperAI

Command Palette

Search for a command to run...

Trained AI Outperforms Biologists in Salmon Lice Detection

Researchers from the Norwegian University of Science and Technology and Wageningen University have developed an artificial intelligence system capable of outperforming trained biologists in detecting salmon louse larvae, a critical advancement for Norway's aquaculture sector. The study, published in Computers and Electronics in Agriculture, addresses a persistent challenge in fish farming: the rapid proliferation of Lepeophtheirus salmonis, a parasite exacerbated by dense net-pen populations that threaten both farmed stock and wild marine ecosystems. Current monitoring methods rely on manual microscopic analysis, a process that is labor-intensive, costly, and prone to human error. In controlled trials, expert biologists required over thirty hours across multiple days to correctly identify 82 percent of louse larvae in complex seawater samples. In contrast, the newly trained AI model completed the same task in thirty minutes, achieving a 97.5 percent identification rate. The system was developed using a custom video microscope that captured more than 120,000 high-resolution images of larvae and surrounding plankton from fjord environments near Ålesund. To overcome data limitations, researchers supplemented the dataset with synthetic imagery generated by rotating, scaling, and adjusting the contrast of original frames, ensuring the model could recognize larvae across varying orientations and developmental stages. The technology directly supports Norway's regulatory traffic light system, which assigns production limits to fish farms based on parasite loads to safeguard wild salmon populations. By enabling continuous, high-throughput environmental monitoring, the AI reduces reliance on indirect lice estimates derived from farmed fish. This shift minimizes uncertainty in parasite mapping, allows for more precise allocation of preventive treatments, and informs strategic decisions regarding farm placement and operational scaling. Lead researcher Lars Christian Gansel emphasized that precise, real-time detection of free-swimming larvae is essential for evaluating the efficacy of current control measures and developing targeted interventions. The AI platform eliminates the need to manually sift through millions of plankton particles, transforming salmon louse surveillance from a sporadic, expert-dependent task into a standardized, data-driven process. As the technology moves toward broader industry deployment, it promises to strengthen environmental compliance, reduce disease transmission risks, and support the long-term sustainability of marine aquaculture.

Related Links