-
Notifications
You must be signed in to change notification settings - Fork 0
Data Sharing Agreement
The following Data Sharing Agreement/ Memorandum of Understanding has been sent to data contributors.
The purpose of the data sharing agreement is to ensure proper coordination, attribution, and best use of the acoustic data and annotations provided by contributors to the 2026 Detection, Classification, Localization and Density Estimation using passive acoustics competition dataset. Duration of Agreement This agreement will commence on MM/DD/YYYY and will be in place indefinitely.
Data to be shared include raw audio data, metadata, and annotations. These may be in the form of short segments (clips) containing training targets (e.g. killer whale calls) or longer segments representative of the overall soundscape in the area they were collected. Metadata will include the recording time, location, sample rate and other available attributes to match ICES recording requirements where possible. https://www.ices.dk/data/data-portals/Pages/Continuous-Noise.aspx Annotations may consist of killer whale ecotypes and ‘other’ classifications. Killer whale annotations will include ecotypes (resident, Biggs, offshore, etc). Other examples may include ship noise and humpback whale presence. Annotations should link to specific files, or file streams (e.g. selection tables) and, where possible, contain the minimum, maximum, frequency and start/end times. Data and annotations will be divided into two sets. The first train/test and the second evaluate. Training and testing dataset will represent annotated audio clips. The evaluation set shall consist of longer (>1hr) file or file streams.
Data will be used as the DCLDE 2026 workshop competition dataset. This will be a publicly available dataset which will allow for the creation and evaluation of detectors and classifiers for killer whale ecotypes. We intend that the data be released with an accompanying descriptive scientific manuscript. All contributing members are encouraged to actively participate in the conception, writing, critical evaluation, and final approval of the paper. All groups contributing data and annotations will be included as authors unless otherwise agreed upon.
Data will be stored on a publicly accessible server such as Box, Dryad, or an institutional database. The details of this will be evaluated closer to the DCLDE 2024 at which point we intend to release the train/test data and part of the evaluation dataset. A subset of the evaluation dataset will be retained by the DCLDE 2026 coordinators in order to evaluate user submissions. Payment No payment is presently available for this work Signatures
TBD. Data should be licensed under open source and permissive (e.g. MIT). Acknowledgment of data contributors is requested. https://en.wikipedia.org/wiki/Open-source_license
- OrcaSound - agreed in principal
- JASCO - agreed in principal
- SMRU Consulting - agreed in principal
- SIMRES
- DFO