Published November 3, 2025 | Version v1.0.0
Dataset Open

TABMON detections dataset

Authors/Creators

  • 1. ROR icon Norwegian Institute for Nature Research
  • 2. Galway-Mayo Institute of Technology
  • 3. University of Amsterdam
  • 4. ROR icon Forest Science and Technology Centre of Catalonia
  • 5. ROR icon Consejo Superior de Investigaciones Científicas
  • 6. ROR icon Centre for Research on Ecology and Forestry Applications
  • 7. ROR icon Pompeu Fabra University
  • 8. ROR icon Université de Toulon

Description

This Zenodo dataset contains the data collected by the TABMON project. This include both the raw detections made by BirdNET and the validation of some of these detections made by expert ornithologists. This Zenodo dataset contains:

1) The raw detections made by BirdNET

The dataset contains a .zip file containing all the BirdNET detections (BirdNET v2.4) inferred from the TABMON dataset. The current version of this dataset contains the detections from January 2025 up to January 2026, later versions will include the latest analysis.

The tabmon_detection.zip file is structured such that there is a folder per country. Inside the country's folder are subfolders for each device (= site) that itself contains .parquet files. The parquet files contains the detections.

The .parquet files can be opened using standard softwares such as R and Python and behave like regular dataframes when loaded. The parquet files are formatted as follow:

filename deployment_id start_time confidence scientific name max uncertainty
2025-06-03T09_38_17.237Z.mp3 20250603_FR_9_22858ce8 216 0.5779327750205994 Asio otus 0.982403877654254
2025-06-03T09_38_17.237Z.mp3 20250603_FR_9_22858ce8 216 0.19663108885288239 Bubo bubo 0.982403877654254
2025-06-03T09_48_22.406Z.mp3 20250603_FR_9_22858ce8 93 0.26167941093444824 Otus scops 0.829270066880144

Note that because of the multilabel nature of BirdNET, there can be multiple classifications for a single segment. This indicates that two bird species may have sang at the same time. We define the column names as:

  • Filename refers to the file that has been analyzed
  • deployment_id links up to the metadata of where the site has been deployed (including habitat type, coordinates ...)
  • start_time refers to the start of the detection of BirdNET v2.4
  • confidence refers to the BirdNET score
  • scientific name refers to the species that has been classified by BirdNET
  • max uncertainty refers to the entropy of the model prediction (for a more in depth description, please refer to Bernard et al.)
 
2) The validation dataset
 
To validate the detections made by BirdNET we hired expert ornithologists and sent them sample clips and their associated BirdNET classification. The experts then confirmed or infirmed the detections. More details about our methodology in Cretois et al., 2026.
 
The tabmon_validation.zip file contains two .zip files (annotation_audio_FR1.zip and annotation_audio_NO1.zip) containing the audio file clips. The two accompanying .csv contain the expert validation and are formatted this way:
 
Filename Model predicted species Model confidence Species you identify Your confidence (Low, Medium, High) Comments
2025-01-16T15_19_52.679Z_889bef22_88.mp3 Cygnus cygnus [0.91834551] Cygnus cygnus High NA
 
To date, 2000 BirdNET detections have been annotated. Moreoever, instead of using excel tables, we are now annotating using the TABMON Listening Lab web application.
 
We would like to thank Tom Østeras who annotated the Norwegian BirdNET detections, and Tanguy Loïs from BioPhonia, who annotated the French BirdNET detections.
 
References
 
 
Bernard, C., McEwen, B., Cretois, B., Glotin, H., Stowell, D., & Marxer, R. (2025). Data-driven Sampling Strategies for Fine-Tuning Bird Detection Models. bioRxiv, 2025-10.
Cretois, B., Rosten, C., Wiel, J., Barile, C., McEwen, B., Bernard, C., ... & Sethi, S. (2026). TABMON–real-time acoustic biodiversity monitoring across Europe.
 

Files

tabmon_detections.zip

Files (1.6 GB)

Name Size Download all
md5:13dfe4654288f41544269e1340b70fe7
1.4 GB Preview Download
md5:852cf4ce89d11e71bc4163926686c5ef
204.3 MB Preview Download