Two projects are noteworthy this week: one aims to prevent salmons from escaping in rough seas, and the other to get rid of lice on farmed salmons. The commonality of both projects is that both deal with salmon, and both use artificial intelligence to achieve the goal.
Salmon farms, even those close to the shore, can experience very rough seas. In such situations, salmon can escape the nets and that may have detrimental effects on wild fish. One of the major problems with salmon farmed in high density are diseases those fish might carry, that are not relevant to food safety, but may adversely affect wild fish. One example are parasites like lice. While there are mitigation measures for the farmed fish, they no longer apply to escaped salmon and consequently, such diseases might spread. In addition, escaped salmon might also swim upstream and mingle with wild salmon in the mating process, potentially weakening existing wild populations. Therefore, it is important to control and prevent salmon escapes. Currently, the control is done by remotely operated underwater vehicles that are equipped with cameras and operated by a person in a control room that monitors the images. A tiring exercise. So here, artificial intelligence, correctly trained, can operate independently, without operator support to maintain a specific position, even in rough sea. And moving salmon between the net and the boat in rough seas is a challenge that might be tackled with the help of AI. This is, at least the basis of the project between SINTEF Ocean and SFI Exposed that has developed a laser camera system for so-called net-relative navigation. This system, it is believed, will reduce the number of salmon escapes during rough seas.
The other issue with farmed salmon are sea lice. Sea lice are a major issue for the aquaculture industry and using veterinary drugs is not a sustainable solution. A new, also AI-supported approach, is the use of cleaner fishes, typically ballan wrasse and lumpfish. These eat the lice off the salmons. However, not all individuals from the same species are equally efficient in doing that. So how, among so many fish, do you select the most efficient ones? This is where AI comes into play. A joint project between the University of Stirling, Swansea University and Otterferry Seafish uses deep learning to identify from video images the most efficient lice-eating individuals from ballan wrasse and lumpfish. The character traits – relevant for delousing efficiency – of the fish that will be evaluated include how fish behave when cohabiting, boldness, shyness, social interaction and even aggression. Once the evaluation is completed and evaluated by artificial intelligence, it is anticipated that the best delousers are identified. In a second step, these can be used for future breeding programmes. This essentially leads to producing individuals specialised for delousing and highly efficient on the job.
These are just two examples of how the needs of the fisheries and aquaculture industry change, and it is obvious that the job descriptions need to be adapted. And this is where the EIT Food-funded AGAPE project with its skills platform can help match emerging needs with existing skills, and if skill gaps are present, suggest how to upgrade the existing skillset.