ShrimpWiz

The aim of the project is to record and validate animal welfare and mortality throughout the entire production chain of a shrimp farm using automated image recognition software. The project is linked to a number of agricultural policy objectives such as competitive agriculture, healthy nutrition and safe food.

ShrimpWiz project as a beacon for animal welfare in aquaculture

Shrimp in German supermarkets come almost exclusively from farms outside the EU. Proof of the species-appropriate husbandry of these animals would provide consumers with guidance when shopping and at the same time demonstrate the advantages of domestic, land-based shrimp farming. Due to the high level of turbidity in the pond facilities for aquaculture shrimp that exist worldwide, it is however almost impossible to record animal welfare using artificial intelligence such as automated image recognition, as stressed or already diseased shrimp are difficult to distinguish visually from healthy animals. In contrast to pond production, the water in Oceanloop GmbH's land-based facilities is clear. This advantage in terms of animal welfare and sustainability has already been successfully demonstrated in the predecessor project, which was also funded by the BLE. The length and number of animals could be determined with an accuracy of over 90% and visual stress indicators (also with 90% accuracy) could be recorded for the first time. In the project now being funded, this AI is to be further developed to market maturity.

How can AI lead to higher animal welfare in modern aquaculture?

In modern land-based aquaculture, due to the high stocking densities, it is generally only possible for the system operator to record the number of animals in the stock and derive a demand- based feeding management system from this by regularly capturing, measuring and weighing the animals. Both underfeeding and overfeeding lead to stress and reduced animal welfare due to competition or impairment of water quality. It is practically impossible to examine each individual animal for symptoms of stress or even sick animals during stocktaking - external signs of stress are barely visible to the naked eye, even under optimal lighting conditions. However, the combination of high-resolution image quality, state-of-the-art camera hardware, powerful computers and the latest generation of AI-based image processing models now makes it possible to detect these visual stress symptoms - in real time.

“The short-term goal is market-ready animal welfare software”

Over the next two years, the project partners, the company Oceanloop Kiel GmbH and the Alfred- Wegener- Institute for Polar and Marine Research Bremerhaven, will work together with one of the leading companies in AI development to generate a large amount of image data in order to develop market-ready animal welfare software for land-based shrimp farming.