DAMP-VPB: Digital Archive of Mobile Performances - Smule Vocal Performances Balanced

<p>The DAMP-balanced dataset contains 24874 solo singing performances from 5429 singers singing a collection of 14 songs.<br> The structure of the DAMP-balanced is that the last 4 songs are designed to be the test set, and the first 10 songs<br> could be partitioned into any 6/4 train/validation split (permutation) that the singers in train and validation set sang<br> the same 6/4 songs collections according to the 6/4 split (the number of total recordings for train and validation set<br> are different from split to split, since there are different number of singers that all sang the same 6/4 split for<br> different split).</p> <p>For example, a subset from the dataset splitting the 14 songs into 6/4/4 train/validation/test sets having 276/88/224<br> performances sang by 46/22/56 singers could be extracted from the dataset and be used for machine learning tasks.<br> Each singer in the train, validation and test set, sang each of the 6/4/4 songs once respectively, thus making the<br> collections of performances &quot;balanced&quot; respect to the song collections.</p> <p>List of songs:</p> <ol> <li>one call away - Charlie Puth(3912)</li> <li>say you won`t let go - James Arthur(3255)</li> <li>allof me - John Legend(2856)</li> <li>closer -The Chainsmokers(2873)</li> <li>seven years - Lukas Graham(2942)</li> <li>despacito- Luis Fonsi(1287)</li> <li>more than words - Extreme(586)</li> <li>lostboy - Ruth B.(1183)</li> <li>loveyourself - Justin Bieber(3019)</li> <li>rockabye- Clean Bandit(2737)</li> <li>part of your world - Jodi Benson(56)</li> <li>when I was your man - Bruno Mars(56)</li> <li>chandelier- Sia(56)</li> <li>cups - Anna Kendrick(56)</li> </ol> <p>Users of this dataset must read and accept Smule&#39;s Research Data License Agreement (LICENSE.txt).</p>