Lisna Thomas

https://doi.org/10.64823/ijter.2621008

The quality and freshness of fish and seafood are crucial factors for ensuring their safety, preventing seafood fraud, and sustainable consumption. The latest achievements in computer vision, Internet of Things (IoT), and artificial intelligence technologies allow for automated evaluation of fish quality. Solutions that can identify fish species and predict its freshness at the point of purchase by end-users remain rare. In this paper, we propose FishID+ – a consumer-oriented and portable solution that incorporates deep learning, multi-sensor IoT, and Explainable Artificial Intelligence (XAI) for real-time fish species identification and freshness prediction. For identifying fish species, our system uses convolutional neural network (CNN) trained on smartphone photos. Our IoT system is based on an ESP32 microcontroller and includes sensors for measuring NH₃ concentration, volatile organic compounds (VOCs), temperature, and humidity. The collected information is classified in a lightweight machine learning algorithm that categorizes the fish into Fresh, Consume Soon, or Spoiled state. Moreover, using SHAP-based explanations, the system identifies which of visual and IoT features have a higher contribution in final decision. The results of prediction, the freshness score, and recommendations are shown through a mobile application, enabling consumers to make informed purchasing decisions without requiring laboratory testing. By integrating species recognition and freshness assessment into a single intelligent platform, FishID+ addresses current limitations in existing seafood monitoring systems. The proposed framework is expected to reduce household food waste, minimize foodborne health risks, enhance consumer confidence, and support sustainable food safety through accessible, explainable, and intelligent IoT technologies.