ALEVIA
- TITLE: Data space for the automated classification of fry using computer vision and machine learning for precision and sustainable aquaculture.
- DATE: 2025 - 2026
- FUNDING PROGRAM: Grants in the field of digitalization, for the digital transformation of strategic sectors through the development of technological products and services for data spaces.
- PARTNERS:
Description
ALEVIA addresses one of the most critical challenges in aquaculture: the accurate and early classification of fry. Traditionally, this process is performed manually, which entails high costs, subjectivity, and a risk of errors. To overcome these limitations, ALEVIA proposes a solution based on computer vision, machine learning, and data interoperability, designed to identify deformities in gilthead seabream (Sparus aurata) or European seabass (Dicentrarchus labrax) and assess fish health status in real-time.
The platform integrates Convolutional Neural Networks (CNNs), predictive algorithms, and an image processing system optimized for large data volumes, achieving inferences in less than one second per image. Thanks to its modular and scalable architecture, ALEVIA transforms images into high-value structured data, facilitating predictive analysis, traceability, and the generation of useful knowledge for strategic decision-making in aquaculture production.
Furthermore, the project aligns with European data space architectures (Gaia-X), guaranteeing interoperability, data sovereignty, and regulatory compliance. This makes ALEVIA not only a tool for production optimization but also a pioneering system in precision aquaculture, contributing to environmental sustainability, animal welfare, and the global competitiveness of the sector.
Key objectives
Develop a secure image ingestion system via a web portal and scalable API, capable of handling at least 500 concurrent requests.
Implement an image processing engine to identify deformities in fry with a minimum accuracy of 90% and response times of <1s.
Design a distributed and traceable storage system, with encryption and data retrieval in <100 ms, based on NoSQL databases and cloud storage.
Integrate advanced predictive analytics to forecast growth, viability, and health patterns from large data volumes (>1 TB).
Guarantee interoperability within data spaces through REST API connectors compliant with Gaia-X, featuring federated access control and complete traceability.
Optimize a real-time inference engine, capable of processing up to 1000 concurrent requests with a latency of <500 ms, integrable into automated fish farms.
Feed efficiently, produce more, and reduce costs.
Discover how to improve your fish farm.
Maximize your fish farm’s performance with smart feeding systems that reduce feed waste and improve feed conversion. At Fish Farm Feeder, we offer innovative solutions for all aquaculture production stages.
