Development of prediction system for egg quality of bluefin tuna based on deep learning

2023 Fiscal Year Final Research Report

• Project/Area Number: 20K15587
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 40030:Aquatic bioproduction science-related
• Research Institution: Yokohama City University
• Principal Investigator: Terayama Kei 横浜市立大学, 生命医科学研究科, 准教授 (50789328)
• Project Period (FY): 2020-04-01 – 2024-03-31
• Keywords: 卵質評価 / クロマグロ / 深層学習

Outline of Final Research Achievements

Egg quality evaluation is an important issue in fish seed production. However, research on indices necessary for egg quality prediction in Pacific bluefin tuna has not progressed sufficiently, and few egg quality evaluation methods have been conducted, especially focusing on the apparent morphology. In this study, we worked on the development of a new method for estimating egg quality (normal hatching rate and survival days without feeding) from only egg images immediately after spawning using deep learning. Data on egg images immediately after spawning and hatching were collected, and a prediction model using a convolutional neural network was trained. As a result, the prediction was successfully made with an accuracy of 0.856 for normal hatching rate that exceeded that of aquaculture researchers. Furthermore, a system for egg quality evaluation was also developed for bulk egg images.

Free Research Field

情報科学

Academic Significance and Societal Importance of the Research Achievements

本研究により、太平洋クロマグロの卵質が産卵直後の画像のみから推定可能であることが示された。これにより、卵質が良いと期待される卵のみを利用することでより効率的な種苗生産に繋がることが期待される。また、本研究は、太平洋クロマグロに限らず、深層学習による魚の卵質評価が可能であることを示唆している。他の魚種でも同様の予測モデルを構築することで、高精度な卵質評価が可能になり、より効率的な種苗生産につながることが期待される。