DCASE 2021 Task 5: Few-shot Bioacoustic Event Detection Development Set
Creators
- 1. Queen Mary University of London
- 2. Centre National de la Recherche Scientifique (CNRS)
- 3. University of Konstanz & Max Planck Institute of Animal Behavior
- 4. BIOTOPIA Naturkundemuseum Bayern
- 5. AGH University of Science and Technology
- 6. -
- 7. University of Konstanz
- 8. Cornell Lab of Ornithology
Description
General Description
The development set for task 5 of DCASE 2021 "Few-shot Bioacoustic Event Detection" consists of 19 audio files acquired from different bioacoustic sources. The dataset is split into training and validation Sets.
Multi-class annotations are provided for the training set with positive (POS), negative (NEG) and unkwown (UNK) values for each class. UNK indicates uncertainty about a class.
Single-class (class of interest) annotations are provided for the validation set, with events marked as positive (POS) or unkwown (UNK) provided for the class of interest.
Folder Structure
Development_Set.zip
|_Development_Set/
|__Training_Set/
|___BV/
|____*.wav
|____*.csv
|___HT/
|____*.wav
|____*.csv
|___JD/
|____*.wav
|____*.csv
|___MT/
|____*.wav
|____*.csv
|__Validation_Set/
|___HV/
|____*.wav
|____*.csv
|___PB/
|____*.wav
|____*.csv
Development_Set_Audio.zip has the same structure but contains only the *.wav files.
Development_Set_Annotations.zip has the same structure but contains only the *.csv files
Dataset statistics
Some statistics on this dataset are as follows, split between training and validation set and their sub-folders:
-----------------------------------------------------
TRAINING SET
-----------------------------------------------------
Number of audio recordings | 11
Total duration | 14 hours and 20 mins
Total classes (excl. UNK) | 19
Total events (excl. UNK) | 4,686
-----------------------------------------------------
TRAINING SET/BV
-----------------------------------------------------
Number of audio recordings | 5
Total duration | 10 hours
Total classes (excl. UNK) | 11
Total events (excl. UNK) | 2,662
Sampling rate | 24,000 Hz
-----------------------------------------------------
TRAINING SET/HT
-----------------------------------------------------
Number of audio recordings | 3
Total duration | 3 hours
Total classes (excl. UNK) | 3
Total events (excl. UNK) | 435
Sampling rate | 6,000 Hz
-----------------------------------------------------
TRAINING SET/JD
-----------------------------------------------------
Number of audio recordings | 1
Total duration | 10 mins
Total classes (excl. UNK) | 1
Total events (excl. UNK) | 355
Sampling rate | 22,050 Hz
-----------------------------------------------------
TRAINING SET/MT
-----------------------------------------------------
Number of audio recordings | 2
Total duration | 1 hour and 10 mins
Total classes (excl. UNK) | 4
Total events (excl. UNK) | 1,234
Sampling rate | 8,000 Hz
-----------------------------------------------------
-----------------------------------------------------
VALIDATION SET
-----------------------------------------------------
Number of audio recordings | 8
Total duration | 5 hours
Total classes (excl. UNK) | 4
Total events (excl. UNK) | 310
-----------------------------------------------------
VALIDATION SET/HV
-----------------------------------------------------
Number of audio recordings | 2
Total duration | 2 hours
Total classes (excl. UNK) | 2
Total events (excl. UNK) | 50
Sampling rate | 6,000 Hz
-----------------------------------------------------
VALIDATION SET/PB
-----------------------------------------------------
Number of audio recordings | 6
Total duration | 3 hours
Total classes (excl. UNK) | 2
Total events (excl. UNK) | 260
Sampling rate | 44,100 Hz
-----------------------------------------------------
Annotation structure
Each line of the annotation csv represents an event in the audio file. The column descriptions are as follows:
TRAINING SET
---------------------
Audiofilename, Starttime, Endtime, CLASS_1, CLASS_2, ...CLASS_N
VALIDATION SET
---------------------
Audiofilename, Starttime, Endtime, Q
Open Access
This dataset is available under a Creative Commons Attribution 4.0 International (CC BY 4.0) license.
Contact info
Please send any feedback or questions to:
Veronica Morfi: g.v.morfi@qmul.ac.uk