Automated classification of banknotes is essential for financial security and e-transaction systems. Conventional manual authentication is time-consuming and error-prone, which necessitates automated systems. This article introduces BanglaNotes, a benchmark dataset for Bangladeshi banknote denomination classification. The dataset contains 5073 images of Bangladeshi paper currency of nine denominations (2, 5, 10, 20, 50, 100, 200, 500, and 1000 BDT), with high quality and consistent representation. Every image in the dataset is labeled according to its denomination. The dataset is suitable for training and testing machine learning algorithms for currency classification and it supports research in financial automation and deep learning-based classification algorithms. It also offers a benchmark for developing robust algorithms for banknotes' real-life applications in banks, mobile payment systems, and ATMs. The dataset is publicly released to promote innovation and further research in banknote classification and recognition.