In this study, we have innovatively designed and developed a benchmark dataset OUC-MOI-ID specifically tailored for the task of Marine Organism Individual IDentification (MOI-ID), with a focus on Sea Cucumber (Actinopyga echinites) and Leopard Coral Grouper (Plectropomus leopardus). This dataset lays a critical foundation for our research. Through the utilization of the dataset, we have implemented and benchmarked state-of-the-art methods, including Convolutional Neural Networks (CNNs), Transformers, Fine-Grained Classification methods, and Vision Mamba techniques, to validate the dataset’s effectiveness. Our comprehensive experiments demonstrate the robustness and high-performance capability of the dataset in supporting identification systems. Notably, fine-grained classification networks such as Swin Transformer and Vision Mamba methods exhibited superior performance, especially in underwater scenarios. These findings underscore the value of the dataset and provide a robust benchmark methodology for future research, offering researchers a valuable reference for utilizing this dataset in related studies. Our work highlights the efficacy of the dataset design and establishes a comprehensive benchmark dataset that facilitates enhanced strategies for marine biological research, monitoring, and breeding practices.