Publicly-available Datasets
Below are a number of publicly-available datasets for audio version identification:
| The Covers80 Dataset | A dataset with low quality audio consisting of 160 songs which are split into two disjoint subsets A and B, each with exactly one version of a pair of songs, for a total of 80 pairs. Mostly '80s and early '90s pop music. This was one of the first publicly available datasets used by researchers |
| Da-TACOS | A dataset with pre-extracted features and metadata for 15,000 songs for a "benchmarking subset" and 10,000 songs for a "cover analysis subset." |
| CoversBR | A database with pre-extracted features (similar to Da-TACOS) for 102,298 songs, distributed into 26,366 groups of covers, of mostly Brazilian music. |
| Covers1000 | A dataset of pre-extracted features of 395 groups of songs, along with a live demo of some alignment algorithms. |
| Kara1k Karaoke Songs Dataset | A dataset with features for 2000 songs: 1000 originals and 1000 corresponding karaoke versions. Also a great dataset for singing voice analysis. |
| The SecondHandSongs Dataset | Another dataset based off of annotations from secondhandsongs.com, which is a subset of the Million Songs Dataset consiting of about 20,000 tracks with EchoNest features. |
| The YouTubeCovers Dataset | A collection chroma, CRP, and CENS features for 350 songs of various genres. |
| https://secondhandsongs.com/ | A community project of annotations of cover songs. |