The effectiveness of automated table detection systems is often limited by the limitations of specialized datasets, particularly for non-English documents and scanned pages with rotational artifacts. To address this limitation, we introduce ViFinTab, a new public dataset for table detection in Vietnamese financial statements. A key feature of this dataset is its use of rotated bounding box annotations, which provide superior localization accuracy for tables in skewed or tilted documents. ViFinTab contains a diverse collection of financial tables, specifically income statements, balance sheets, and cash flow statements. In addition to the dataset, we provide a strong deep learning baseline model to demonstrate its utility and establish a benchmark for future research.