Many of the publicly available electrocardiogram (ECG) databases either have a low number of people in the database, each with longer recordings, or have more people, each with shorter recordings. As a result, attempting to split a single database into training, testing, and, optionally, validation datasets is challenging. Some models seem to do well with larger training sets, but that leaves only a small set of data for testing. Moreover, if the ECG is segmented by heartbeat, the data are further limited by the number of heartbeats in the recording.