Plankton are a diverse group of organisms relevant to oceanic ecosystems. Photosynthetic plankton contributes significantly to the oceanic biological carbon pump, influencing the global carbon cycle, and is a foundational component of marine food webs. Given their sensitivity to environmental changes, plankton biomass and diversity are recognized as Essential Ocean Variables. Studying plankton at meaningful scales is challenging due to the ocean’s vastness and complexity, the wide range in plankton size, and their intricate distribution patterns. Traditional analysis relies on intensive and time demanding manual classification. While digitizing samples has helped, the large volume of image data still requires extensive manual annotation. Therefore, computer assisted image classification is essential for advancing oceanographic research. This paper explores the use of deep learning techniques for classifying plankton images. Good classification results were achieved using transfer learning approaches. However, Convolutional Neural Networks (CNNs) trained from scratch did not perform well, likely due to the limited number of available samples. This underscores the importance of transfer learning when dealing with small datasets. Among the evaluated architectures, ResNet50V2 consistently demonstrated the best performance across all configurations, including those with and without regularization techniques like data augmentation and early stopping. This positions ResNet50V2 as the most robust and effective architecture for this specific classification task. Importantly, these findings confirm the viability of classifying plankton from digital images, a crucial step towards automating marine biodiversity monitoring.