This dataverse contains data (radiographs) that were used during the machine learning training process of identifying Retained Foreign objects (RFOs) in radiographs. The dataverse also serves as a data repository for radiologists to access to assist their clinical judgements. This research aims to enhance the decision-making of radiologists by developing a machine learning model to reduce the occurrence of RFOs. Every data in the RFO dataverse follows a consistent naming convention: the source of the radiograph, the type of surgical/foreign object, the location of the RFOs, the angle that the radiograph was taken from, the orientation, the surgical procedure, and the gender of the patient. However, some information associated with certain data cannot be obtained due to a lack of information and technological limitation. Please contact the creator of this dataverse for any inquiry pertaining to the sources of radiographs in the dataverse.