AI technology helps count sea lice

Photo of author
Written By Chris

New expertise is getting used to rely sea lice on salmon extra precisely, resulting in a higher understanding of learn how to breed resistance to the parasite.

Benchmark Genetics Norway has collaborated with the Danish Technological Institute on a brand new course of that makes use of imaging expertise and synthetic intelligence to analyse the variety of sea lice in actual time. Counting this manner is way sooner than the guide counting method usually used.

As Hooman Moghadam, Senior Analysis Scientist with Benchmark, explains, counting the variety of sea lice is vital to assessing the fish’s resistance and, due to this fact, helps to construct up datasets of the genetic make-up (“phenotypes”) that may finest assist salmon to outlive.

The imaging expertise makes use of a mix of a half-circular light-dome (CSS dome mild HPD2-400FC) and a 5-megapixel monochrome digicam to take a picture of every salmon. The high-power light-dome has three particular person triggered color diodes: crimson (622 nm), inexperienced (525 nm) and blue (470 nm).

The mono-camera takes one image at every wavelength, after which pictures are post-processed utilizing the developed AI. It was discovered that this imaging system supplies the consumer with the very best visibility and distinction between the lice and the fish. The algorithm makes use of a deep studying segmentation mannequin primarily based on a a number of convolutional community structure U-net picture mannequin, initially developed for biomedical picture segmentation.

The AI mannequin is educated to phase the ocean lice and salmon. After picture segmentation, the ocean lice are filtered and counted. By repeating this course of many instances and routinely augmenting the picture look in varied methods, the mannequin learns the form and turns into sturdy towards various fish and lice sizes, picture angles, illumination variations and many others. The coaching dataset is created by manually marking pixels with lice within the picture.

The mannequin can then examine its outcomes towards the coaching dataset and will get educated. The larger the coaching dataset, the extra correct the algorithm can detect lice. The mannequin’s efficiency enchancment is verified by preserving a subset of the annotated information for mannequin validation.

The Danish Technological Institute (DTI) is a self-owned and non-profit analysis and expertise organisation.

Categories AI

Leave a Comment