A Sustainable Deep Learning based Computationally Intelligent Seafood Monitoring System for Fish Species Screening

Abstract

Seafood being one of the major source of omega-3 fatty acid, is widely consumed around the world, thus at the same time affects baby-growth due to high mercury. Therefore, an automatic and intelligent seafood classification system is demanded greatly. This study proposed a shallow but effective deep learning based computationally intelligent system that can classify nine different fish species. The model is trained with publicly available data set, called A Large Scale Fish Dataset. 35% of data is used for testing while rest is reserved for training and validation. Experimental results shows that proposed model achieved an overall accuracy of 94%, thus surpasses many previously proposed models. Detail analysis shows that the model secured better F1-score (98%) on Red Mullet fish.