VOICe Dataset

<p>VOICe: A novel dataset for the development and evaluation of generalizable sound event detection domain adaptation methods!<br> <br> VOICe consists of 1449 different mixtures of three different sound events (&quot;baby crying&quot;, &quot;glass breaking&quot;, and &quot;gunshot&quot;):</p> <ul> <li>1242 mixtures with background noise of three different categories of acoustic scenes (&quot;vehicle&quot;,&quot; outdoors&quot;, and &quot;indoors&quot;), mixed under 2 SNR values (-3, -9 dB), that is 207 mixtures x 3 acoustic scenes x 2 SNRs = 1242</li> <li>&nbsp;207 mixtures without any background noise.</li> </ul> <p>&nbsp;VOICe is offered for sound event detection domain adaptation from one acoustic scene to another, or between sound events with background noise and without background noise.<br> <br> You can also find more information about the dataset in our paper: <a href="https://www.google.com/url?q=https://arxiv.org/pdf/1911.07098.pdf&amp;sa=D&amp;source=hangouts&amp;ust=1578483524111000&amp;usg=AFQjCNFfL1pv0_wXhnpNPEktej3tK_eqIw">https://arxiv.org/pdf/1911.07098.pdf</a></p>