Data Delicacies: Endemic Fish as New Dataset for Culinary Artificial Intelligence

Endemic fish species are key components in seafood culinary excursions. Despite the increasing interest in leveraging technology to enhance various seafood culinary activities, there is a shortage of comprehensive datasets containing images of seafood used in artificial intelligence research, mainly those showcasing endemic fish. This research endeavors to bridge this gap by increasing the accuracy of fish recognition and introducing a new dataset comprising images of native fish for application in various machine-learning investigations. The dataset consists of three categories of native fish: rabbit fish (149 images), squid (144 images), and red snapper (234 images). SqueezeNet is used for feature extraction, followed by the data partitioning into training and test sets at a 60:40 ratio. Several classifier algorithms were employed to construct models from the dataset, exclusively utilizing 2-fold cross-validation and providing modeling accuracy results of 96.9% for the Fish4knowledge dataset and 99.7% for the Endemic Fish primary dataset.