Underwater inspections are essential for maintaining offshore infrastructure but face challenges such as high collection costs, low image quality, and limited labeled data. Due to the diversity of objects and structures, traditional models are limited, making techniques that learn from few examples particularly relevant and representative. This work presents an image classifier for underwater inspections based on the One-Shot Learning approach. This technique enables machine learning models to recognize new classes from a very limited number of examples, overcoming the common limitation of scarce datasets in underwater applications. A Siamese neural network was employed to learn a similarity function between image pairs. The model was trained on a dataset of underwater inspection images containing different categories, such as ROVs, pipelines, objects, junctions, and equipment, and evaluated in N-way tasks using metrics including accuracy, precision, recall, and F1-score. The results demonstrated that the approach effectively identifies unseen classes with competitive performance, highlighting the applicability of One-Shot Learning in underwater inspection scenarios with limited data availability. This study contributes to the development of intelligent tools that can assist in the main tenance and monitoring of submerged structures.