Computer Vision and Artificial Intelligence based Salmon Identification and automated long-term welfare assessment in aquaculture net-pens

The cAIge project is transforming how we monitor and assess salmon in fish farms by exploiting advanced computer vision and artificial intelligence (AI) methods to automate welfare monitoring at the level of individual fish. Using continuous stereo-video recordings from commercial cages, cAIge develops algorithms and tools that can extract relevant information directly from large volumes of video data. The project is organized into four work packages: Work Package 1: Automatic feature extraction. We are developing deep-learning models that detect and track high-level features of salmon, such as eyes, fins, body contours, and external conditions (wounds, scale loss, deformities). These detections form the basis for quantifying health indicators at scale. Work Package 2: Individual re-identification. By combining biometric features and advanced tracking, we aim to re-identify individual salmon across time. This enables following the same fish over weeks and months, providing unique insights into individual trajectories of growth and welfare in large cage populations. Work Package 3: Linking welfare with environment and operations. We integrate fish-level observations with environmental data (e.g., temperature, oxygen) and farm operations (e.g., treatments, crowding). AI reasoning strategies are used to analyse how these factors contribute to fish welfare and to predict welfare outcomes under different conditions. Work Package 4: Validation and biological relevance. All developed methods are validated against expert assessments and operational data to ensure biological accuracy and practical applicability. Field trials and experimental studies are central to this validation. cAIge advances the use of AI in aquaculture by enabling objective, automated, and scalable monitoring of salmon welfare. The project contributes to improved fish health, better decision support for farmers, and more sustainable production practices. The project has also been endorsed as a UN Ocean Decade project, highlighting its role in applying artificial intelligence to promote sustainable aquaculture and support the UN Sustainable Development Goals.

Technology development in cAIge will allow automated, machine-assisted monitoring of fish health and welfare in aquaculture farms at hitherto unprecedented precision levels. This is to be achieved by leveraging advanced AI-based computer vision technology along with individual fish re-identification algorithms for uninterrupted welfare assessment over time. In addition, the relationship between observed welfare status, environmental factors, and farming practices will be integrated into comprehensive models, thereby gaining a better understanding of how these factors interact to shape optimal holding conditions of fish in aquaculture. We emphasize that monitoring the welfare and responses to operational and environmental factors of individual fish over a prolonged period of time represents a significant advancement over the current practice to monitor the fish, thereby enabling researchers in the future to assess fish welfare and behavior on a scale never experienced before.