Motif Mining and Unsupervised Representation Learning for BirdCLEF 2022

TL;DR

This paper presents an unsupervised learning approach using spectrogram motifs and triplet loss to classify bird sounds for the BirdCLEF 2022 challenge, achieving competitive results.

Contribution

It introduces an unsupervised representation learning method based on spectrogram motifs and triplet loss for bird sound classification.

Findings

Achieved a score of 0.48 on the public leaderboard

Developed an unsupervised spectrogram motif embedding

Demonstrated effectiveness of triplet loss in bird sound classification

Abstract

We build a classification model for the BirdCLEF 2022 challenge using unsupervised methods. We implement an unsupervised representation of the training dataset using a triplet loss on spectrogram representation of audio motifs. Our best model performs with a score of 0.48 on the public leaderboard.