Soundscape Datasets for Few-Shot Bird Sound Classification

<p>This repository provides easy access to open-source soundscape datasets of bird sounds, specifically optimized for few-shot classification.</p> <p><code>soundscapes.zip</code> contains evaluation soundscape datasets from the BIRB benchmark (https://arxiv.org/abs/2312.07439), downsampled to 16kHz, preprocessed using CNN14 from PANNs (https://arxiv.org/abs/1912.10211), to select a 6-second window with the highest bird activation, and converted to Pytorch (.pt) format to facilitate usability for evaluating deep neural networks.&nbsp;</p> <p>These preprocessed datasets are employed in the work "<em>Domain-Invariant Representation Learning of Bird Sounds</em>" (https://arxiv.org/abs/2409.08589), which evaluates the few-shot learning capabilities of deep learning models trained on focal recordings (e.g., Xeno-Canto) and tested on soundscape recordings.</p> <h2>Dataset Structure</h2> <h3>Validation Dataset</h3> <ul> <li><strong>POW (</strong><code>pow.pt</code><strong>):&nbsp;</strong>The validation dataset consists of 16,047 examples across 43 classes and is organized as a dictionary with&nbsp;<code>'data'</code> and<code> 'label'</code> keys representing bird sounds and their corresponding labels. Storing the entire validation dataset in a single tensor enables rapid loading and efficient processing, significantly accelerating the validation process. Classes with only one example are removed, as they are insufficient for one-shot classification tasks. Source: https://zenodo.org/records/4656848#.Y7ijhOxudhE</li> </ul> <h3>Test Datasets&nbsp;</h3> <p>Each test dataset is structured with multiple subfolders, each labeled with an eBird species code to represent data for a specific bird species.</p> <ul> <li><strong>SSW (</strong><code>ssw/</code><strong>):</strong> Contains 50,760 examples across 96 classes. Source: https://zenodo.org/records/7079380#.Y7ijHOxudhE</li> <li><strong>NES (</strong><code>coffee_farms/</code><strong>):</strong> Contains 6,952 examples across 89 classes. Source: https://zenodo.org/records/7525349#.ZB8z_-xudhE</li> <li><strong>UHH (</strong><code>hawaii/</code><strong>):</strong> Contains 59,583 examples across 27 classes. Source: https://zenodo.org/records/7078499#.Y7ijPuxudhE</li> <li><strong>HSN (</strong><code>high_sierras/</code><strong>):</strong> Contains 10,296 examples across 19 classes. Source: https://zenodo.org/records/7525805#.ZB8zsexudhE</li> <li><strong>SNE (</strong><code>sierras_kahl/</code><strong>):</strong> Contains 20,147 examples across 56 classes. Source: https://zenodo.org/records/7050014#.Y7ijWexudhE</li> <li><strong>PER (</strong><code>peru/</code><strong>):</strong> Contains 14,768 examples across 132 classes. Source: https://zenodo.org/records/7079124#.Y7iis-xudhE</li> </ul> <p>Code and detailed instructions, including data loading, model implementation, and few-shot evaluation, can be found at: https://github.com/ilyassmoummad/ProtoCLR</p>