•  
  •  
 

Abstract

Fish freshness classification is critical for protecting public health and ensuring efficient economic, regulatory and environmental sustainability. Classifying accurately reduces the risk of foodborne illness, protects product quality, builds consumer trust and supports sustainable resource conservation through waste minimization. However, the traditional methods for determining fish freshness are variable, time consuming and subjective, precluding practical use. This research presents an improved framework that integrates image data fusion and a deep learning ResNet model to differentiate fresh and nonfresh fish. From multiple sources, a comprehensive dataset including 16,640 samples was curated, and data fusion was used to increase the diversity and reliability of the extracted features. The classification model was developed via ResNet, which is well known for its extraordinary feature extraction features. High performance was shown by the proposed approach, with 92% precision, 94% recall and an F1 score of 0.93 for fresh fish. For nonfresh fish, the precision was 95%, the recall was 93%, and the F1 score was 0.94. Overall accuracy of classification. This suggests that the model proposed in this work is a feasible and reliable solution for real-time fish freshness classification that outperforms traditional methods. This is followed by the use of image data fusion along with ResNet to further state deep learning in food quality assessment, maintain environmental sustainability, contribute to public health, and improve economic value. This research illuminates the value of data fusion for enhancing model performance while offering a novel means to address central problems in the seafood industry.

Download

Share

COinS