Abstract

Fish recognition and classification represent significant challenges in marine biology and agriculture, promising fields for advancing research. Despite advancements in real-time data collection, underwater fish recognition and classification still require improvement due to challenges such as variations in fish size and shape, image quality issues, and environmental changes. Feature learning approaches, particularly utilizing convolutional neural networks (CNNs), have shown promise in addressing these challenges. This study focuses on video-based fish species classification, employing a feature learning-based extraction method through CNNs. The process involves two main stages: detection and classification. To address the detection and classification in video a Faster Region Convolutional Neural Network (RCNN) with transfer learning techniques are applied, achieving a mean average precision of 84% for detection and classification tasks. These techniques offer promising avenues for enhancing fish recognition and classification in diverse environments