Transfer Learning with Semi-Supervised Dataset Annotation for Birdcall Classification

TL;DR

This paper explores transfer learning combined with semi-supervised dataset annotation to improve birdcall classification, leveraging existing models and feature engineering to enhance performance in a competitive setting.

Contribution

It introduces a novel approach using BirdNET and MixIT for dataset annotation and transfer learning in birdcall classification tasks.

Findings

Effective classification of African bird species achieved

Semi-supervised annotation improves model performance

Transfer learning reduces need for extensive labeled data

Abstract

We present working notes on transfer learning with semi-supervised dataset annotation for the BirdCLEF 2023 competition, focused on identifying African bird species in recorded soundscapes. Our approach utilizes existing off-the-shelf models, BirdNET and MixIT, to address representation and labeling challenges in the competition. We explore the embedding space learned by BirdNET and propose a process to derive an annotated dataset for supervised learning. Our experiments involve various models and feature engineering approaches to maximize performance on the competition leaderboard. The results demonstrate the effectiveness of our approach in classifying bird species and highlight the potential of transfer learning and semi-supervised dataset annotation in similar tasks.