The National Library of Medicine's Pillbox dataset contained 8,693 photographs of pills, with an accompanying database of drug information. It was built to help with the identification of unknown pills.