<![CDATA[Assessing the quality of tuna loin remains a pivotal aspect of the global seafood industry, necessitating precise, consistent, and efficient grading methods that can be broadly implemented. This study addresses these challenges by developing a robust, cloud-native system for automated tuna loin quality classification. Utilizing a tailored image dataset, the system's core processing is handled by a scalable cloud-based backend on Google Cloud Platform, specifically employing Cloud Run for serverless inference. The deep learning model, EfficientNetV2M, is optimized into the ONNX format and executed efficiently by ONNX Runtime within this cloud environment, achieving a classification accuracy of 96% with rapid prediction times. An intuitive Flutter frontend application serves as the user interface, facilitating the transmission of image data to the cloud service and displaying real-time grading results. This architectural design ensures dynamic resource allocation, high availability, and cost-effectiveness through a pay-per-use model. Data integrity and security are maintained via HTTPS for secure communication between the frontend and the cloud-deployed backend. The integration of Docker for containerization, Google Cloud Run for serverless deployment, and Flask for API management collectively yields a highly scalable, reliable, and efficient system. This research presents a robust, cloud-centric solution for automated tuna loin quality classification, offering real-time predictions, secure data handling, and a user-friendly interface suitable for industrial quality control and research applications. Keywords — cloud computing, serverless, Google Cloud Run, Docker, ONNX, deep learning, computer vision, real-time prediction.]]>
Copyrights © 2025
