A Prototypical Triplet Loss for Cover Detection

TL;DR

This paper introduces a new triplet loss and multi-pitch input representation to improve cover detection accuracy in music datasets, especially under realistic conditions with few covers per work.

Contribution

It proposes a novel prototypical triplet loss and multi-pitch input data to enhance cover detection performance in practical scenarios.

Findings

Significant accuracy improvements in large dataset lookup.

Enhanced live song identification performance.

Effective clustering of cover tracks with the new loss.

Abstract

Automatic cover detection -- the task of finding in a audio dataset all covers of a query track -- has long been a challenging theoretical problem in MIR community. It also became a practical need for music composers societies requiring to detect automatically if an audio excerpt embeds musical content belonging to their catalog. In a recent work, we addressed this problem with a convolutional neural network mapping each track's dominant melody to an embedding vector, and trained to minimize cover pairs distance in the embeddings space, while maximizing it for non-covers. We showed in particular that training this model with enough works having five or more covers yields state-of-the-art results. This however does not reflect the realistic use case, where music catalogs typically contain works with zero or at most one or two covers. We thus introduce here a new test set incorporating these constraints, and propose two contributions to improve our model's accuracy under these stricter conditions: we replace dominant melody with multi-pitch representation as input data, and describe a novel prototypical triplet loss designed to improve covers clustering. We show that these changes improve results significantly for two concrete use cases, large dataset lookup and live songs identification.