An accurate count would lead to a greater understanding of how to breed resistance to sea lice in salmon.

Last week it was reported that a recent study that looked at the state of sea lice in the Pacific Ocean and found that the parasite is increasingly becoming resistant to a pesticide known as SLICE, or emamectin benzoate, one of the main tools that the salmon farming industry uses to combat the menace. An assessment usually involves manually counting the sea lice on salmon, a process that is often erroneous and sometimes accompanied by deliberate under-reporting. However, Benchmark Genetics, Norway, has collaborated with the Danish Technological Institute on a new, AI-enabled process that uses imaging technology and artificial intelligence to analyse the number of sea lice in real time. (The Fish Site)

How the AI works

The AI algorithm has been trained to tell the sea lice apart form the salmon host. It also learns, through a repetitive process, to identify and count the sea lice against varying fish and lice sizes, image angles, illumination variations etc.

The algorithm uses a deep learning segmentation model, based on a multiple convolutional network architecture U-net image model, initially developed for biomedical image segmentation.

Now, each salmon is photographed in real time by a 5 megapixel monochrome camera while being illuminated by a high-power, half-circular light-dome (CSS dome light HPD2-400FC) that has three individual triggered colour diodes: red (622 nm), green (525 nm) and blue (470 nm). The camera takes a picture of the salmon at each of these wave-lengths.

These images are processed by the AI model to deliver the results.

According to Hooman Moghadam, Senior Research Scientist with Benchmark, counting the number of sea lice is critical to assessing the fish’s resistance and, therefore, helps to build up datasets of the genetic make-up (“phenotypes”) that can best help salmon to survive.

Related Story: The Norwegian AI Lab Helps Salmon Aquaculture

Image: Kind courtesy of Prince William Sound Science Center