eduzhai > Applied Sciences > Engineering >

Neural Loop Combiner Neural Network Models for Assessing the Compatibility of Loops

  • king
  • (0) Download
  • 20210507
  • Save

... pages left unread,continue reading

Document pages: 8 pages

Abstract: Music producers who use loops may have access to thousands in loop libraries,but finding ones that are compatible is a time-consuming process; we hope toreduce this burden with automation. State-of-the-art systems for estimatingcompatibility, such as AutoMashUpper, are mostly rule-based and could beimproved on with machine learning. To train a model, we need a large set ofloops with ground truth compatibility values. No such dataset exists, so weextract loops from existing music to obtain positive examples of compatibleloops, and propose and compare various strategies for choosing negativeexamples. For reproducibility, we curate data from the Free Music Archive.Using this data, we investigate two types of model architectures for estimatingthe compatibility of loops: one based on a Siamese network, and the other apure convolutional neural network (CNN). We conducted a user study in whichparticipants rated the quality of the combinations suggested by each model, andfound the CNN to outperform the Siamese network. Both model-based approachesoutperformed the rule-based one. We have opened source the code for buildingthe models and the dataset.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×