Employment of Artificial Intelligence in Fish Fraud Identification

Resumo

Acquiring fish for consumption whether in restaurants, supermarkets or fishmongers can become a
frustrating task from the moment it is perceived that the species at hand is not the expected fish. Added to this is
the fact that fraud always aims to deceive to make a profit. There are some applications that aim to identify fish
species; however, they have no specific focus on certain species. With this, the user applies to all species of fish
and ends up disappointed with the inefficiency of the application. This project aims to implement an application
that uses artificial intelligence to identify certain species of fish. The main point of the application are fish species
and cannot be used for ornamental fish. The method used was to obtain images and configuration of models of
artificial neural networks for image recognition. During the tests the implemented model reached 75% accuracy.
The model had difficulty recognizing three species accurately, it was the case of Merluza fish that was confused
with Salmon and Trout. It is believed that this occurred due to the low number of samples, but in short, the model
achieved satisfactory results.