Printed-and-Scanned Document Pages

500 printed-and-scanned A4 document pages with full structural ground truth. Each page is generated from a Wikipedia article, rendered in one of four layout templates (article, memo, report, two-column) and one of ten fonts, then printed on a Konica Minolta bizhub C550i and scanned on the flatbed of the same device at 200, 300, 400, and 600 DPI. The same pages are additionally scanned on a consumer HP DeskJet 2630 at 300 DPI as a lower-cost reference surface. Every page carries embedded fiducials (a QR code and four AprilTag corners) so that scans can be aligned back to the original render at sub-pixel precision. Per-page JSON labels describe layout, font, page geometry, word boxes in both points and pixels, and the source Wikipedia article (title, URL, revision ID). Suitable for OCR benchmarking, layout-analysis evaluation, scan-quality and DPI ablation studies, fiducial-based document alignment and registration research, and synthetic-to-real domain shift work.