Fish recognition system with elements of machine learning

Abstract

This paper is wrotten by Khalid Sabih, student of group KH-417f in the field of computer science in Lviv Polytechnic National University. The subject area of the paper is fish recognition system with machine learning elements. This paper examines the need of machine learning systems in the fishing industrie, and for what they are needed. Fishing as an industrie is one of the biggest one out there with loads of fishermen that needs loads of information. The object of research is to provide useful and important information for fishermen. The subject of the research is image recognition systems and data classification. Research method is a transfer learning algorithms. This work is focusing more towards the fishermen that are amateurs and doing it just for fun, and telling them they can make profit of it, adding to that it can be as a side job for them. As an amateur it is hard to know all kind of fish species and how much money the costs. That is the reason behind this paper to offer them this kind of information with easy steps. Using transfer learning through VGG16 and with the right data, it was able to creat a model with 94.18%. For a prototype was Streamlit a great tool for that. With all of this it was possible to creat Catchfish which is an image recognition system that allows amateur fishermen with a scan and a click to have all the information needed about the fish catched.