Fisheries monitoring programmes are essential to ensure the sustainable management of marine resources. However, current electronic monitoring systems—based on cameras, GPS and sensors—still require operators to manually review thousands of hours of video to identify species and quantify catches.
This process is costly, time-consuming and difficult to scale. To move towards more efficient, transparent and near real-time control, it was necessary to develop deep learning–based solutions capable of automatically identifying captured species and estimating their proportions and sizes in each fishing set.
AZTI has developed a computer vision model designed to automate the classification and measurement of the main species caught by large tropical tuna freezer vessels—yellowfin (Thunnus albacares), skipjack (Katsuwonus pelamis) and bigeye (Thunnus obesus)—operating in the Atlantic, Indian and Pacific oceans.
The process includes:
Video preprocessing and standardisation, correcting for variations in angle, lighting and vessel-specific conditions.
Creation of a dedicated training dataset for tropical tunas, with thousands of images collected from conveyor belts in the fishing deck.
Development of a deep learning model capable of segmenting fish in images, classifying them by species and measuring their size.
Automatic conversion of measurements into species proportions and weight estimates per fishing operation.
Although challenges remain—such as glare, lens dirt or overlapping fish—the project already incorporates improvements like stereoscopic cameras, which enable 3D reconstruction and measurement of fish parts that are not directly visible.
The first operational computer vision model applied on Spanish tropical tuna freezer vessels, capable of identifying and measuring tropical tuna species.
Automatic estimation of species proportions and sizes per set, dramatically reducing reliance on manual video review.
A solid technological foundation for future advanced electronic monitoring systems that are more robust and scalable.
An active roadmap towards more accurate 3D models, enabling application in scientific monitoring and fisheries control.
This project represents a decisive step towards more efficient, transparent and objective data-driven monitoring, strengthening the sustainability of one of the world’s most important tuna fleets.
Sectors: Fisheries and aquaculture sector, Technology 4.0
Research lines: Blue Economy, Digitalisation, Efficient, sustainable fisheries and aquaculture
Research sublines: Sustainable fishery management