To train a supervised algorithm, Cascade (code repository), which infers spike rates from calcium signals, we have recorded and collected a large database of ground truth recordings. The database consists of 27 single datasets (including the GENIE datasets, some of the Spikefinder datasets, the Allen Institute datasets, and several datasets that we recorded ourselves), covering 7 different brain regions, 8 different calcium indicators, two different species, totalling more than 38 hours of recording. More details about the datasets are available in a Resource Article published with Nature Neuroscience.