KhushbuParmar313/Birds_audio_files

Bird Audio Dataset for Bird Species Classification

Overview

This dataset contains bird vocalization recordings collected from the Xeno-Canto database, a global community-driven repository of bird sounds.

The data collection process initially targeted a broad set of bird species across multiple countries and continents to obtain diverse recordings of bird songs and calls. Following collection, the dataset was curated to improve class balance and suitability for machine learning applications.

To ensure that each class contained a sufficient number of training samples, only species with at least 500 audio recordings were retained in the final dataset. After this filtering process, the dataset consists of 105 bird species (classes).

The resulting dataset provides a balanced and diverse collection of bird vocalizations suitable for species classification, bioacoustic research, biodiversity monitoring, and deep learning applications.

Dataset Creation Pipeline

1. Bird recordings were collected from Xeno-Canto using scientific species names and country-based queries.
2. Metadata associated with each recording was retrieved and preserved.
3. Species with fewer than 500 recordings were excluded from the final release.
4. The remaining recordings were organized into species-specific folders.
5. The final dataset containing 105 species was split into training, validation, and test sets.

Dataset Split

The dataset is divided into three non-overlapping subsets:

• Training Set: 70%
• Validation Set: 15%
• Test Set: 15%

The split was performed on a per-species basis to maintain similar class distributions across all subsets.

Dataset Summary

• Source: Xeno-Canto (https://xeno-canto.org)
• Number of Classes: 105 bird species
• Number Recordings per Class: 500
• Data Type: Bird songs and calls
• Train/Validation/Test Split: 70% / 15% / 15%
• Geographic Coverage: Recordings collected from multiple countries across Europe, Asia, Africa, North America, South America, and Oceania
• Primary Applications:
  • Bird species classification
  • Bioacoustic analysis
  • Environmental sound recognition
  • Deep learning research
  • Biodiversity monitoring
  • Wildlife conservation