The steelpan is a pitched percussion instrument that although generally known by listeners is typically not included in music instrument audio datasets. This means that it is usually underrepresented in existing data-driven deep learning models for fundamental frequency estimation. Furthermore, the steelpan has complex acoustic properties that make fundamental frequency estimation challenging when using deep learning models for general fundamental frequency estimation for any music instrument. Fundamental frequency estimation or pitch detection is a fundamental task in music information retrieval and it is interesting to explore methods that are tailored to specific instruments and whether they can outperform general methods. To address this, we present SASS, the Steelpan Analysis Sample Set that can be used to train steel-pan specific pitch detection algorithms as well as propose a custom-trained deep learning model for steelpan fundamental frequency estimation. This model outperforms general state-of-the-art methods such as pYin and CREPE on steelpan audio - even while having significantly fewer parameters and operating on a shorter analysis window. This reduces minimum system latency, allowing for deployment to a real-time system that can be used in live music contexts.