Enhancing Fish Quality Inspection through Yolov9 Based Deep Learning Techniques

Quality of fish products is important not only for safety concerns but also for general food standards. Classical methods of inspection are highly subjective, with high human error and hence are considered to be inefficient and unreliable. This paper presents the use of YOLOv9, the advanced deep learning model, to assess the freshness of fish products based on some key indicators: eye transparency and gill color, with images captured by a mobile phone camera in RGB. The model was trained, tested, and validated with a full dataset consisting of day-wise as well as hour-wise images of the eyes and gills of Rohu fish along with images of other species. CNN-LSTM Hybrid Model: A hybrid combination of the CNN's strengths in feature extraction with long short-term memory network utilities for sequence analysis further improved the accuracy. This way, the inspection of fish becomes of the highest quality, with minimal human errors and better standards of food safety at all levels of supply.