Plankton play a critical role in aquatic ecosystems, contributing to oxygen production, nutrient cycling, and the regulation of global carbon dynamics. Effective monitoring of plankton populations is essential for understanding environmental change and ecosystem health. However, traditional plankton classification relies on manual image annotation by taxonomic experts, a process that is labor-intensive and difficult to scale. These challenges are particularly acute in data-sparse regions such as the Caribbean, where labeled datasets are scarce. This study presents a plankton classification approach for a microscopy image dataset collected from the coastal waters of Trinidad and Tobago by the Department of Life Science at the University of the West Indies (UWI). Due to the limited size of this regional dataset, transfer learning was applied using the National Data Science Bowl (NDSB) dataset, which includes over 30,000 images across 121 plankton classes. Convolutional neural networks (CNNs) using the NDSB data are fine-tuned to adapt to the local Caribbean samples. The proposed method significantly improves classification performance, especially in the context of limited data and class imbalance. The results demonstrate the viability of using large, publicly available datasets to enhance local ecological monitoring efforts, offering a scalable and efficient alternative to manual annotation in underrepresented regions.