Knowledge of marine fish species is critical to marine ecosystems and the fishing industry. The three fish species focused on in this project are Polyprion americanus, Trigloporus lastoviza, and Anthias anthias. Polyprion americanus, known as wreckfish, is often found in the deep sea and has high economic value. Trigloporus lastoviza, or red gurnard, is usually found in sandy or muddy waters and is famous for its wing-like fin shape. Meanwhile, Anthias anthias, better known as swallowtail seaperch, is a fish that lives on coral reefs and is known for its bright colors. Accurate and rapid identification of these species is crucial in conservation, scientific research and marine resource management.
https://www.kaggle.com/datasets/giannisgeorgiou/fish-species
This project utilized an Artificial Neural Network-based image classification technique to identify the three seawater fish species. By utilizing Conv2D layers, batch normalization, pooling, and global average pooling, and using the SGD optimizer, the model was able to achieve 88.13% accuracy on the test data. The trained model was then deployed using TensorFlow JS and accessed through a web application running on a local server using the “Web Server for Chrome” extension. The results of this project show that the model can effectively and efficiently classify fish images into three specified species, providing a practical solution for real-time applications in research and the fishing industry.