Sovereign 8s.is mirror of standardgalactic/BirdSoundClassif
Find a file
Repository files (latest commit first)
Filename Latest commit message Latest commit date
2025-11-07 16:34:43 +01:00
demo_results Add demo 2024-02-16 10:06:19 +01:00
model_weights Update weights 2025-11-06 09:56:22 +01:00
nbm_model Update inference 2025-11-06 09:54:06 +01:00
xc_utils Delete xc_utils/dist directory 2024-12-04 19:09:36 +01:00
.gitattributes Update weights 2025-11-06 09:55:49 +01:00
.gitignore Update inference 2025-11-06 09:54:06 +01:00
bird_dict.json FRCNNv2 2024-12-04 19:03:00 +01:00
README.md Update README.md 2025-11-07 16:34:43 +01:00
requirements.txt FRCNNv2 2024-12-04 19:03:00 +01:00

NBM: an Open Dataset for the Acoustic Monitoring of Nocturnal Migratory Birds in Europe (Link to paper).

image info

NBM dataset on Zenodo

The NBM dataset can be found and downloaded at this address: (https://zenodo.org/records/17544346).

NBM object detection model

🐥 Installation

  • Install python version 3.10
  • Create virtual environment with conda (or venv) and activate it. Here is how to do with conda.
conda create venv
conda activate venv
pip install -r requirements.txt

🐔 Model weights

The model_weights directory includes both the checkpoints and the configuration file.

🦉 Prerequisite

To run the model you will need the checkpoint and config files, along with the dictionnary file (bird_dict.json) linking species names to ids.

🐦 Model execution

Execute the following command line instruction to analyze all .wav files in a directory:

python nbm_model/nbm_detect.py --ckpt model_weights --audio_dir <PATH_TO_WAV_FILES_DIRECTORY>

The NBM object detection model currently classifies (fligt) calls from common European songbird species. Please refer to the paper for an exhaustive list of the main species. A number of species lacking substantial training data can still be identified but with reduced accuracy.