The ACMUS-MIR data set is a selection of music taken from the Músicas Regionales Archive in Medellin, Colombia. The data set currently contains a total of 624 segments of annotated music. The data set was compiled to support the development of computational methods for four main tasks: 1) Recognition of simple and compound meters, 2) Instrumental format recognition, 3) Scale detection, and 4) Vocal music classification. For this reason, the ACMUS-MIR data set currently comprises four independent sets (one for each task): the rhythm set, the instrumental format set, the scale set, and the vocal format set. For more information about our project, please visit our website and github. For additional datasets from the ACMUS project, see: Sesquialtera in the Colombian bambuco: Perception and estimation of beat and meter