Classification of Fish Freshness using YOLOv7

This study uses innovative computer vision techniques to give a thorough review of the freshness assessment of milkfish and tilapia. The study focuses on classifying fish species and assessing freshness using the YOLOv7 algorithm based on physical characteristics, including the state of the skin and eyes. A dataset with 2940 hand-collected photos was used for model training and evaluation, using different methods such as data preprocessing, augmentation, and model construction. Findings showed the model's strong ability to identify different fish species and assess freshness parameters, with a noteworthy mean Average Precision (mAP@50) score of 0.0.88. The model's performance in various classes and freshness categories was examined in detail using precision-recall analysis, revealing opportunities for development and improvement. The study also demonstrated the critical importance of morphological parameters, finding a strong positive correlation between fish freshness and skin and eye condition. Future research should prioritize ethical considerations, improve model accuracy and speed, integrate multisensory data, and investigate other morphological aspects for freshness evaluation. All things considered, this study offers a dependable framework for classifying fish and assessing their freshness, with implications for improving sustainability and quality management in the seafood industry.