et al., T. (2025). Web-Based Deep Learning Model for Tuna Loin Quality Assessment in the Fisheries Processing Industry. Egyptian Journal of Aquatic Biology and Fisheries, 29(2), 1181-1207. doi: 10.21608/ejabf.2025.419360
Tupan et al.. "Web-Based Deep Learning Model for Tuna Loin Quality Assessment in the Fisheries Processing Industry". Egyptian Journal of Aquatic Biology and Fisheries, 29, 2, 2025, 1181-1207. doi: 10.21608/ejabf.2025.419360
et al., T. (2025). 'Web-Based Deep Learning Model for Tuna Loin Quality Assessment in the Fisheries Processing Industry', Egyptian Journal of Aquatic Biology and Fisheries, 29(2), pp. 1181-1207. doi: 10.21608/ejabf.2025.419360
et al., T. Web-Based Deep Learning Model for Tuna Loin Quality Assessment in the Fisheries Processing Industry. Egyptian Journal of Aquatic Biology and Fisheries, 2025; 29(2): 1181-1207. doi: 10.21608/ejabf.2025.419360
Web-Based Deep Learning Model for Tuna Loin Quality Assessment in the Fisheries Processing Industry
The Maluku region recognizes tuna loin as a premier processed commodity derived from capture fisheries resources. This study developed an innovative non-destructive quality prediction model for tuna loin utilizing deep learning technology. Traditionally, quality assessment of tuna meat relies on color and texture evaluation through organoleptic/sensory methods, which demands significant time and specialized expertise. Expanding on prior CNN-based research for tuna treatment classification (Tupan et al., 2025); this investigation pioneers the application of Deep Convolutional Neural Networks (DCNN) specifically for tuna grading purposes. The research focused on evaluating the effectiveness of various CNN architectures (ResNet, DenseNet, and Inception) for tuna loin grade classification, while simultaneously developing an integrated prediction system available as both a web-based application and Android mockup. Performance analysis of the multi-architecture CNN algorithms revealed varied accuracy levels across different grade classification schemes. In the three-tier classification system (Alpha, Bravo, and Charley), DenseNet demonstrated superior performance with 94.64% accuracy, while ResNet achieved 91.07% and Inception reached 83.93%. These results highlight the significant potential of deep learning approaches for automated quality assessment in fishery products. The resulting integrated platform features an intuitive user interface that enables tuna loin image uploads for analysis by the dual-classification prediction model. Validation testing confirms the successful implementation of both treatment and grading classification systems, providing the fish processing industry with a comprehensive, efficient tool for real-time quality assessment. The deployed solution addresses critical industry challenges including consistency in quality evaluation and enhanced decision-making capabilities throughout the tuna processing supply chain.
View on SCiNiTO
Top