Sign in to view source links and access this dataset
Description
On the night of September 23, 2015, six autonomous recording units captured 20,000 ten-second audio clips near Ithaca, New York. This dataset is a curated subset of the BirdVox-full-night recordings, with 10,017 clips (50.09%) containing at least one bird vocalization. It was created as a development set for the DCASE Bird Audio Detection challenge, facilitating the development and evaluation of bioacoustic classification models.
Use Cases
[Training bird audio detection models] based on [10,017 labeled positive examples of bird vocalizations].
[Benchmarking machine learning algorithms] based on [the standardized 10-second clip format from the DCASE challenge].
[Studying nocturnal bird migration patterns] based on [audio recordings captured during a single fall night in 2015].
[Evaluating sensor performance] based on [recordings from six distinct autonomous recording units (ROBIN sensors)].
Strengths
Contains a precise count of 20,000 audio recordings, with 10,017 (50.09%) explicitly annotated as containing bird vocalizations.
Provides detailed provenance, including the exact date (September 23, 2015), location (Ithaca, NY), and hardware (six ROBIN sensors).
Serves as a benchmark dataset for a recognized challenge (DCASE Bird Audio Detection), ensuring relevance for comparative research.
Limitations
Metadata completeness is low; column names, file formats, and the exact data structure are not specified in the provided sources.
The dataset's temporal coverage is extremely narrow, consisting of recordings from a single night, which may limit generalizability.
There is a minor version conflict in the sources, with both Version 1.0 and Version 2.0 listed for March 2018.
Provenance
Source
Cornell Lab of Ornithology, New York University (Center for Urban Science and Progress, Music and Audio Research Lab)
Collection Method
Recordings captured by six ROBIN autonomous recording units.
Time Range
September 23, 2015
Freshness
March 2018
Geography
Ithaca, New York, USA
Dataset is a derivative work of the BirdVox-full-night dataset. License is listed as 'Open Access (green)'.