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

TL;DR

This paper introduces neural network models, including CNN and Siamese architectures, to automate the assessment of loop compatibility in music production, outperforming existing rule-based systems.

Contribution

It presents the first dataset for loop compatibility, compares neural network architectures, and demonstrates CNN's superior performance over Siamese networks and rule-based methods.

Findings

CNN outperforms Siamese network in user ratings

Both models outperform rule-based compatibility estimation

Open-sourced dataset and code for reproducibility

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 to reduce this burden with automation. State-of-the-art systems for estimating compatibility, such as AutoMashUpper, are mostly rule-based and could be improved on with machine learn-ing. To train a model, we need a large set of loops with ground truth compatibility values. No such dataset exists, so we extract loops from existing music to obtain positive examples of compatible loops, and propose and compare various strategies for choosing negative examples. For re-producibility, we curate data from the Free Music Archive.Using this data, we investigate two types of model architectures for estimating the compatibility of loops: one based on a Siamese network, and the other a pure convolutional neural network (CNN). We conducted a user study in which participants rated the quality of the combinations suggested by each model, and found the CNN to outperform the Siamese network. Both model-based approaches outperformed the rule-based one. We have opened source the code for building the models and the dataset.