Automated Fish Measurement and Classification Using Convolutional Neural Networks (CNNs)

Abstract

Managing fisheries requires regular monitoring and assessment of fish populations. Traditional methods of evaluating fish stocks, particularly their size, can be time-consuming, labor-intensive, and inaccurate. Recently, digital image processing (DIP) and machine learning (ML) have emerged as promising technologies to automate fish measurement and classification. In this study, we aim to develop deep learning models to predict, and classify shape and size of the fish using convolutional neural networks (CNNs) and DIP techniques. The study utilizes publicly available fish datasets and evaluates the efficiency of the proposed models using metrics such as precision, recall, and F1 score. The developed models utilize Python programming language with TensorFlow and Keras libraries. The regression component investigates the intricate relationship between various physical attributes of fish, uncovering the connections between body length, height, and weight. This analysis provides valuable insights into the correlations among these attributes, enhancing our understanding of fish characteristics. Simultaneously, the classification segment introduces an innovative approach to fish classification, incorporating shape and size attributes. Through a combination of classifiers and ensemble learning with stacking, exceptional accuracy is achieved in identifying distinct fish classes. This integration of techniques facilitates a more nuanced classification process, allowing for comprehensive categorization based on visual attributes. Our study establishes a robust framework for fish analysis and classification, Utilizing the combined strengths of digital image processing (DIP) and machine learning (ML). The developed models not only enhance the accuracy and efficiency of size classification but also contribute to the broader goal of sustainable fisheries management. This research sets a foundation for future endeavors in automating fish stock assessments, contributing to the advancement of fisheries science and management practices.