Gunshot sound files and spectrogram images from passive acoustic monitoring data in Vietnam.

Introduction

This dataset was created to support benchmarking of automated gunshot detection models using 'torch for R'. The data includes recordings collected from Chu Mom Ray National Park in Vietnam combined with an open dataset of gunshots from Belize (Katsis et al. 2022). The dataset is organized into several folders containing .jpg images and their corresponding .wav audio clips. 

Data Summary

The dataset is divided into three main categories based on the region and use case:

1. imagesvietnamunbalanced: Contains spectrogram images and audio data collected from Chu Mom Ray National Park in Vietnam. These data are used for training and evaluating automated gunshot detection models.
2. imagesvietnam_belize: Includes spectrogram images and audio data from both Vietnam and Belize. The data in this folder is used for benchmarking model performance across different geographical regions.
3. testdatacombined: Separate test data to evaluate performance with a large number of noise clips, representing real-world automated detection scenarios.

Each .jpg image in these directories is associated with a .wav file representing a corresponding audio clip. The .wav files were recorded using passive acoustic monitoring and clips were isolated using manual annotations in Raven Pro Software.

If used please cite: 

Vu, T. T., Phan, D. V., Le, T. S., & Clink, D. J. (2024). Investigating hunting in a protected area in Southeast Asia using passive acoustic monitoring with mobile smartphones and deep learning. Ecological Indicators. 

Vu, T. T., Phan, D. V., Le, T. S., & Clink, D. J. (2024). Gunshot sound files and images from passive acoustic monitoring data in Vietnam. [Data set]. Zenodo. https://doi.org/10.5281/zenodo.13893977

Katsis, Lydia; Hill, Andrew; Piña-Covarrubias, Evelyn; Prince, Peter; Rogers, Alex; Doncaster, C. Patrick; Snaddon, Jake (2022), “Tropical forest gunshot classification training audio dataset”, Mendeley Data, V3, doi: 10.17632/x48cwz364j.3