Deep Learning and Image Processing Approach for Fish Freshness Detection

The proposed paper represents a new method for fish freshness detection based on deep learning and image processing. Our approach is an effective, accurate and Non-Invasive tool to classify fish quality using Convolutional Neural Networks (CNNs). Our model was trained on a diverse dataset that included images of fish at different levels of freshness. This CNN was highly accurate and reliable in distinguishing fresh from spoiled fish. The proposed approach uses convolutional neural networks and transfer learning with MobileNetV2 over ImageNet resulting in an impressive accuracy of nearly 97.50% on independent test data, better than the traditional approaches The proposed system is designed to overcome challenges encountered by conventional inspection techniques and introduces a fast, automatic in-line process for quality assurance. This paper helps decision- making in food safety and quality control of seafood, which is considered as a key advance for computerized fish freshness detection.