This Zenodo dataset contains data products generated by the TABMON project, a transnational passive acoustic monitoring network for birds across Europe. TABMON combines autonomous acoustic sensors, automated species identification with BirdNET, standardized deployment metadata, taxonomic information, and expert validation to support large-scale biodiversity monitoring. This Zenodo dataset contains standardized BirdNET detections datasets, an index of acoustic recordings, deployment metadata, a taxonomic reference table, and an expert validation dataset, together with accompanying documentation and reproducible workflows. The BirdNET prediction datasets, deployment metadata, and taxonomic reference table can be considered as the TABMON EBV-useable datasets (Cretois et al. 2026, https://doi.org/10.1111/2041-210x.70308). 1) The BirdNET detections The tabmon_detections.zip file contains the dataset of BirdNET predictions (BirdNET v2.4) inferred from the TABMON recordings with confidence>0.01. The current version of this dataset contains predictions from January 2025 up to April 2026. Later versions may include additional monitoring periods and updated analyses. The tabmon_detections.zip file is structured such that there is a .parquet file per country that contains the predictions. The .parquet files can be opened using standard software such as R and Python and behave like regular data frames when loaded. The parquet files are formatted as follows: 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 or more bird species may have vocalized at the same time. We define the column names as: • filename refers to the file that has been analyzed. The filename contains the date and start time of the 5-minute recording file in UTC. • deployment_id links to the deployment metadata describing where and when the sensor was deployed. • start_time refers to the start of the BirdNET v2.4 prediction (start of the 3-second segment in the 5-minute recording). • confidence refers to the BirdNET confidence score. • scientific name refers to the species classified by BirdNET. • max uncertainty refers to the entropy of the model prediction (for a more in-depth description, see Bernard et al., 2026, https://doi.org/10.1121/10.0043947). A lighter version of the dataset, tabmon_detections_light.zip, includes only predictions with confidence>0.1, reducing the total size from 8.1 to 1.7 GB. The files_index.parquet file lists all audio files recorded in TABMON, including those with no retained BirdNET predictions. It is useful to know which periods were recorded in order to distinguish species non-detections from periods without acoustic recordings. It should be noted that our aim was to collect continuous recordings 24 hours a day (288 5-minute files per day), but problems in the field (equipment, connectivity, power supply via solar panels) led to significant gaps in the data. files_index.parquet is formatted as follows: filename deployment_id 2025-06-03T09_38_17.237Z.mp3 20250603_FR_9_22858ce8 2025-06-03T09_43_29.112Z.mp3 20250603_FR_9_22858ce8 2025-06-03T09_48_22.406Z.mp3 20250603_FR_9_22858ce8 2) The deployment metadata The Deployment_metadata.ZIP archive contains the public-release deployment metadata file site_info_public.csv together with an accompanying README file. The metadata provide the spatial, temporal, technical, and environmental context for TABMON acoustic sensor deployments across Spain, the Netherlands, France, and Norway. Each row represents an individual deployment, rather than necessarily a unique monitoring site or device. The current file contains 141 deployment records. The number of deployment records exceeds the number of monitoring sites because devices were sometimes replaced or redeployed and a new deployment record was created. The unique DeploymentID is the primary identifier linking the deployment metadata to the BirdNET prediction datasets, where the corresponding field is deployment_id. This allows BirdNET predictions to be associated with the relevant monitoring cluster, site, sensor device, deployment period, geographic location, habitat, and other deployment information. The public-release metadata include: • Country – country in which the sensor was deployed. • DeploymentID – unique identifier for an individual sensor deployment. • Cluster – TABMON monitoring cluster. • Site – name or label of the individual monitoring site. • deploymentBeginDate and deploymentBeginTime – date and UTC time at which the deployment began. • deploymentEndDate and deploymentEndTime – date and UTC time at which the deployment ended, where available. • Latitude and Longitude – geographic coordinates of the deployment in decimal degrees. • DeviceID – identifier of the individual Bugg acoustic recording device. • Microphone_height – microphone height above ground in centimetres. • Microphone_direction – direction in which the microphone was oriented. • Habitat – broad habitat category assigned to the deployment site. For the public release, the following sensitive or site-specific fields were excluded: Coordinates_uncertainty, GPS_device, Adresse e-mail, and Comments. The original latitude and longitude coordinates were rounded to three decimal places, corresponding to an approximate spatial precision of about 100 m in latitude. This preserves the spatial context needed for ecological analyses while avoiding publication of the original high-precision device coordinates. The deployment protocol and metadata collection are described in detail in Cretois et al. (2026), including the accompanying appendices and Supporting Information. The README included in Deployment_metadata.ZIP provides additional descriptions of the metadata fields and public-release modifications. 3) The taxonomic reference table The Taxonomic_reference_table.zip archive provides the reproducible workflow and associated files used to create the TABMON taxonomic reference table. The archive contains the latest R script (Create_taxonomic_reference_table_05.R), the required input taxonomy files, the output file taxonomic_reference_table.csv, and an accompanying README file documenting the input data, workflow, quality-control procedures, and output fields. The taxonomic reference table links the BirdLife/European Breeding Bird Atlas 2 (EBBA2)-based TABMON candidate species list to the BirdNET taxonomy (v0.2-Jun2026). The TABMON candidate species list geographically filters the full BirdNET prediction space using occurrence information from EBBA2 and the Military Grid Reference System (MGRS) coordinates encompassing TABMON monitoring sites. This reduces the BirdNET prediction space from 6,522 to 480 candidate species considered relevant to the TABMON monitoring network (Cretois et al., 2026, https://doi.org/10.1111/2041-210x.70308). Candidate species are first linked to BirdNET through direct scientific-name matching. Where the BirdLife/EBBA2 scientific name does not directly correspond to the BirdNET scientific name, the workflow uses BirdNET scientific_name_aliases as a fallback. Quality-control checks ensure that all 480 candidate species are linked unambiguously to unique BirdNET species identifiers. The resulting taxonomic_reference_table.csv contains: • birdnet_id – unique BirdNET species identifier. • birdlife_scientific_name – BirdLife scientific name used in the EBBA2-based TABMON candidate species list. • scientific_name – scientific name used in the BirdNET taxonomy; this may differ from the BirdLife scientific name when an alias was required for matching. • common_name – English common name used in the BirdNET taxonomy. • match_type – method used to establish the taxonomic link: direct_scientific_name or scientific_name_alias. The taxonomic reference table provides a reproducible crosswalk between the geographically filtered TABMON candidate species list and the BirdNET taxonomy used in the prediction datasets. Together, the standardized BirdNET prediction datasets, deployment metadata, taxonomic reference table, and accompanying documentation comprise the TABMON EBV-useable datasets. These standardized data products provide the basis for subsequent assembly of EBV-ready datasets through additional taxonomic harmonization, expert validation, and calibration of classifier performance. Within TABMON Work Package 3, these datasets support showcase analyses for terrestrial bird species distributions and terrestrial bird migration phenology. The underlying continuous acoustic recordings are not included in this Zenodo repository because of their very large data volume and associated data-management requirements. However, a subset can be made available for specific analysis by contacting benjamin.cretois@nina.no or corentin.bernard@lis-lab.fr. 4) The annotated dataset To validate the predictions made by BirdNET, we hired expert ornithologists and sent them sample clips and their associated BirdNET detections. The experts then confirmed or rejected the predicted species identification, and reported any additional species present. More details about the TABMON methodology are provided in Cretois et al. (2026, https://doi.org/10.1111/2041-210x.70308). The tabmon_annotations.zip file contains three .zip files (annotation_audio_FR1.zip, annotation_audio_NO1.zip, annotation_audio_ES1.zip) containing the audio file clips. The three accompanying .csv files (annotation_table_FR1.csv, annotation_table_NO1.csv, annotation_table_ES1.csv) contain the expert annotations and are formatted as follows: Filename Model predicted species Model confidence Species identified by the annotator Annotator confidence Comments 2025-01-16T15_19_52.679Z_889bef22_88.mp3 Cygnus cygnus [0.91834551] Cygnus cygnus High NA 2025-05-14T07_22_00.817Z_c1405178_61.mp3 Sturnus vulgaris, Sturnus unicolor [0.474231 0.49551758] Sturnus vulgaris, Curruca melanocephala, Himantopus himantopus High NA The guide_for_annoators.pdf file contains the protocol and instructions for the annotators.TABMON subsequently expanded and standardized expert validation using the TABMON Listening Lab web application. We would like to thank Tom Østeras, who annotated predictions from Norway, Tanguy Loïs from BioPhonia, who annotated predictions from France and Lluís Brotons, from the Centre de Ciència i Tecnologia Forestal de Catalunya (CTFC) and the Institut Català d'Ornitologia (ICO), who annotated predictions from Spain. References Bernard, C., McEwen, B., Cretois, B., Glotin, H., Stowell, D., Marxer, R., 2026. Data-driven sampling strategies for fine-tuning bird detection models. The Journal of the Acoustical Society of America 159, 4891–4903. https://doi.org/10.1121/10.0043947 Cretois, B., Rosten, C.M., Wiel, J., Barile, C., McEwen, B., Bernard, C., Boom, M.P., Bota, G., Brotons, L., Serrano-Davies, E., Glotin, H., Kissling, W.D., Marxer, R., Pérez-Granados, C., Stowell, D., Villero, D., van Zweden, J.S., Sethi, S., 2026. TABMON: Design and deployment of a transnational passive acoustic monitoring network for European birds. Methods in Ecology and Evolution 17, 1867–1879. https://doi.org/10.1111/2041-210x.70308

Details This year's BioDCASE Workshop will be held online on the 23rd of October. The program will include invited speakers on the topic of bioacoustics as well as poster sessions of technical reports. The BioDCASE workshop will be hosted the week before the DCASE workshop on the 28-29th of October …

Join us at the University of Salford and be part of Team Salford and one of the UK's fastest growing universities!

Temporal context may aid automated bat-species classification, but the contribution of specific features remains unclear. We investigated whether variation in the interpulse interval (IPI)-the time between consecutive call onsets-provides species-discriminative information and whether transformer-based models are more sensitive to this information than convolutional neural networks. We created two matched datasets from European bat recordings: a natural-IPI condition retaining the original call timing and a normalized-IPI condition in which call onsets were spaced at 50-ms intervals. EfficientNet-B0 and PaSST were fine-tuned and evaluated within each condition. In an additional experiment, each architecture was trained separately on natural-IPI and normalized-IPI recordings, and evaluated on the same natural-IPI test set. Finally, the pretrained classifiers BatDetect2 and BAT were evaluated on both conditions. Within-condition IPI normalization had model-dependent effects. PaSST accuracy differed little between the natural-IPI ($71 \pm 2.3\%$) and normalized-IPI ($70 \pm 6.3\%$) conditions, whereas EfficientNet accuracy increased from $47 \pm 4.7\%$ to $57 \pm 3.9\%$. PaSST exceeded EfficientNet under both conditions. In the cross-condition evaluation, models trained on natural-IPI recordings outperformed those trained on normalized-IPI recordings on the natural-IPI test set: accuracy decreased from 54% to 50% for EfficientNet and from 65% to 57% for PaSST. BatDetect2 and BAT differed little between IPI conditions. Overall, we found limited support for the hypotheses that natural IPI variation contributes substantially to bat-species classification and that it is used more effectively by transformer-based than CNN-based models. Nevertheless, the cross-condition performance decrease shows that results obtained under normalized conditions may not transfer fully to natural recordings.

Limited expert annotation capacity is a pervasive constraint in biodiversity monitoring. Passive acoustic recorders and camera traps generate data faster than experts can analyse them. Machine learning (ML) models can process these data at scale, but their reliability depends on the quality, quantity, and coverage of labelled samples, so expert time remains a constraint. Active learning (AL) eases this bottleneck by selecting, under a fixed annotation budget, the samples expected to improve a model most, and published evidence shows it can reduce the labels needed to reach a target performance. Monitoring programmes, however, face a broader question: how should a limited expert budget be divided so that model training, validation, and the ecological estimates built on model outputs all remain reliable? Because AL selects samples non-randomly, its labels are unsuitable for validation, calibration, or threshold selection, a tension rarely acknowledged. We synthesise AL research across acoustic and image modalities and identify gaps and opportunities. Most studies evaluate query strategies on pre-labelled benchmarks with simulated annotators; deployments in real monitoring workflows are rare and concentrate on birds and cetaceans. Bats, insects, amphibians, and fish are underrepresented, and multimodal applications remain largely unexplored. Evaluation centres on headline reductions in annotation effort, often without random-sampling baselines, per-class results, or calibration analysis, and rarely accounts for the labels required for validation. We provide a tutorial treatment of the AL loop that makes these budget decisions explicit, and a roadmap towards AL methods that support label-efficient training, validation, and trustworthy downstream ecological inference.

De Nederlandse overheid maakt een eigen digitale werkomgeving op basis van de Nederlandse Linux-distributie NixOS. De ontwikkelaars geven een kijkje in de keuken.

As animal behaviorists increasingly use AI in their research, they grapple with how to wield the technology responsibly

Wil je als onderzoeker bijdragen aan het behoud van biodiversiteit en de toekomst van onze leefomgeving? Dan is deze kans bij Naturalis in Leiden voor jou!

After acts of sabotage brought a halt to many train routes in the central, eastern, and northern portion of the Netherlands, Dutch national railway NS said it will "gradually" begin passenger rail service on Tuesday afternoon. Earlier in the morning, debris was placed on the tracks to disrupt rail switches, and trick signal lights to shine red, ProRail confirmed.

Enabling users to monitor the energy usage and carbon footprint of their HPC use.

Large Language Models (LLMs) are machine learning (ML) models that have an increasingly large carbon footprint through their development and use. Efforts to increase the energy efficiency of these models have not translated into reduced consumption due to rebound effects such as Jevons Paradox - that increased efficiency drives increased use. There is therefore a need for additional measures to solve this problem. We suggest that one possible way forward is to use life cycle thinking, and view LLMs as products that can become waste. With this perspective, we investigate the potential of the waste hierarchy from the EU's Waste Framework Directive, which suggests five different measures for how to manage waste: prevention, reuse, recycling, recovery, and disposal. We examine how these measures can inform and motivate new types of thinking and approaches to reducing LLM waste and their environmental impact in general. Applying the waste hierarchy to LLMs highlights that preventing waste is essential for reducing the models' environmental impact, mainly because it reduces the need for training new models. Prevention can be achieved through many existing methods for reusing, "recycling", and "recovering" LLMs. Additionally, disposal can be important both for saving energy and for keeping a considerate attitude to the resources being spent on training LLMs. We also call to attention that prevention of unnecessary use of LLMs carry huge potential for lowering the climate impact of the models.

Made with Tally, the simplest way to create forms.

Independent, indie, electronic, etc - all from The Netherlands and fresssh

PyTorch port of Google Perch v2 bioacoustic backbone - frozen embeddings, linear probing, and deep fine-tuning. - bghani/perchv2-pytorch