We've released a public dataset for document detection to support research and transparency. This dataset is available on GitHub for anyone to use.