Fish classification and deep learning

• Main Author: Zhang, Dawei
• Other Authors: Alex Chichung Kot
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/168085
• Institution: Nanyang Technological University

Nowadays, more and more aquariums prefer to use scientific and technological methods to perform fish classification rather than identification relying on manual. The biggest problem is how to get a good dataset for training and to find a reliable method to identify fish categories with high accuracy. Most of the traditional methods focus on the pretreatment of dataset and physical condition of hardware. While this essay aims to perform fish classification by a unified meta framework for fine-grained recognition. In data collection part, we include 806 images from aquarium dataset and 1608 images from public dataset. After pretreatment of dataset and cropping operation based on YOLOv7, high classification accuracies for tank A, B, C, and D with 90.14%, 81.74%, 92.81%, and 80.35% are achieved for this meta framework.