Tropical forest gunshot classification training audio dataset

DATA SOURCE LOCATION
Data were collected in tropical forest sites in central Belize. Data were recorded in Tapir Mountain Nature Reserve (TMNR) and the adjoining Pook’s Hill Reserve in Cayo District, Belize [17.150, -88.860] and Manatee Forest Reserve (MFR) and surrounding protected areas in Belize District, Belize [17.260, -88.490].

FOLDERS
The folders contain audio files recorded between 2017 and 2021. The ‘Training data’ folder and the ‘Validation data’ folder contain two temporally distinct datasets, which can be used for model training and validation. The training folder consists of 80% of the total dataset, and the validation folder comprises the remaining 20%. Within each of these folders are two folders labelled ‘Gunshot’ and ‘Background’.

FILES
The folders contain 749 gunshot files and over 35,000 background files. The files are in Waveform Audio File Format (wav), and are each 4.09 seconds long. The first 8 alphanumeric characters of the file name corresponds to the UNIX hexadecimal timestamp of the time of recording. Some files contain additional alphanumeric characters after these initial 8 characters, which were used as unique identifying numbers during processing and do not convey any additional information.