딥러닝 기반 실시간 어류 탐지 알고리즘 구현

Abstract

This study proposes a deep learning-based approach for real-time fish detection under underwater environments. By integrating model compression techniques and transfer learning into a YOLOv3 framework, the system aims to achieve efficient and practical detection performance even on resource-constrained platforms. The proposed approach emphasizes adaptability to underwater imaging challenges and potential deployment in marine monitoring systems. This work contributes to the field by presenting a scalable framework for applying deep learning to real-world aquatic ecological monitoring. Furthermore, the proposed method offers potential applications in low-power semiconductor and display-based embedded systems for real-time visual detection.