Investigating the Performance Differences for Different Bird Songs in BirdNet

Investigating the Performance Differences for Different Bird Songs in BirdNet

This project is based on BirdNet, an existing open source birdsong classification tool. There have been studies about the different performances on classification of different bird species. The folders and files contain the original code files from the official BirdNet repositories as well as adjustments made for evaluating and improving results on certain species classification.

The changes will mainly include additionall files added to the folder and changes of the original scripts.

Context and methodology

• Research domains: Data Science, Ornithology
• Purpose: Evaluation and possible improvement of a model according to specific use case
• Created by addition and changes to an existing implementation

Technical details

• The structure will be mainly kept the same as for the original publication. 
• The code is Python code. Therefore an Installation of python and all necessary packages is needed
• Before using this repository it is advised to also look at the original one: https://github.com/kahst/BirdNET

Further details

• Especially retraining a CNN model can be very computationally expensive. Also all the data reused for the project are very large since they are a huge number of audio recordings.