eduzhai > Applied Sciences > Engineering >

Solos A Dataset for Audio-Visual Music Analysis

  • Save

... pages left unread,continue reading

Document pages: 6 pages

Abstract: In this paper, we present a new dataset of music performance videos which canbe used for training machine learning methods for multiple tasks such asaudio-visual blind source separation and localization, cross-modalcorrespondences, cross-modal generation and, in general, any audio-visualself-supervised task. These videos, gathered from YouTube, consist of solomusical performances of 13 different instruments. Compared to previouslyproposed audio-visual datasets, Solos is cleaner since a big amount of itsrecordings are auditions and manually checked recordings, ensuring there is nobackground noise nor effects added in the video post-processing. Besides, itis, up to the best of our knowledge, the only dataset that contains the wholeset of instruments present in the URMP cite{URPM} dataset, a high-qualitydataset of 44 audio-visual recordings of multi-instrument classical musicpieces with individual audio tracks. URMP was intented to be used for sourceseparation, thus, we evaluate the performance on the URMP dataset of twodifferent source-separation models trained on Solos. The dataset is publiclyavailable at this https URL

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...