FRESQO POC
The FRESQO project originated from years of collaboration and friendship, demonstrating the power of interdisciplinary science. It began with eight core executives from the ATHENA Research Center, FORTH-IACM, HCMR, and the Department of Fisheries and Aquaculture, sparked by a Blue Economy datathon in Crete in June 2017. Building on results from prior European projects, the "seed" was planted for FRESQO: Using a hyperspectral camera, artificial intelligence, and data science to determine the freshness of fish.
FRESQO's main objective is the quantitative and objective estimation of fish freshness, further supporting and complementing existing grading methods and control instruments. Conventional calibration in fisheries and aquaculture is typically species-specific, requiring an extensive series of experimental measurements and verifications for each fish species and harvesting area (origin). Consequently, existing instruments often perform well only for a limited range of species
FRESQO effectively addresses this limitation by offering an infrastructure that can provide a highly accurate freshness indication—currently for nine primary commercial species—regardless of their catch area. The designed methodology and infrastructure can be applied to any fish species without direct contact or destruction of the examined fish, which is a key comparative advantage over existing methods.
The FRESQO system leverages color changes in fish tissues using image analysis tools to provide a reliable estimate of the fish freshness index. To achieve this, the latest advancements in hyperspectral image analysis technology were utilized, combined with machine learning models and data science tools for efficient, multi-level data management. This technological approach is grounded in biological-ichthyological research regarding the parameters that define the freshness of fish catches. Specifically, the system utilizes color variations in certain fish body tissues (gills, eye cloudiness) within the invisible spectrum to perform hyperspectral recording and analysis of image color shifts resulting from fish deterioration. The analysis results were supported by parallel chemical analyses and organoleptic measurements to certify the accuracy of the estimations. These accurate estimations were then used by machine learning tools to optimize the final predictions. The final result is a general-purpose, portable, user-friendly, and extremely fast analytical tool/infrastructure that provides a reliable freshness index for all fish catches.
The most significant problem we faced was the long delivery times for mechanical equipment, particularly for foreign imports. This was managed by selecting temporary alternative components and adjusting both the device and the algorithm accordingly.
The Proof of Concept program essentially helped us demonstrate that the solution we developed—with support from the EPAL 2014-2020 project—could meet market terms and requirements. The SCIENCE AGORA GOOGLE POC gave us the enthusiastic boost and support, taking into account market demands (portability and system user-friendliness), to move forward with steady and rapid steps toward creating a spin-off and entering both the national and international markets.
