This repo releases the data and code for the paper:
- MCAD-MM: A Benchmark Dataset and Method for Multi-Channel Acoustic Detection of Marine Mammals
The MCAD-MM introduces a novel approach for the acoustic detection of marine mammals using deep learning. It establishes a comprehensive benchmark dataset and methodology that encompasses a wide range of marine mammal acoustic signals and ambient seawater noises. This provides essential data support and methodological guidance for advancing research in related fields.
The MCAD-MM dataset can be downloaded via the Baidu Netdisk or Hugging Face links.
For installation and data preparation instructions, please refer to PREPARATION.md.
To start using the benchmark method, follow the steps outlined in GETTING_STARTED.md.
The results are provided in MODEL_ZOO.md.
The data and code in this repo are released under the Apache 2.0 license.
The benchmark data of this study is derived from the DCLDE Oahu passive acoustic dataset obtained through NOAA PIFSC's 2017 Hawaiian Islands Cetacean and Ecosystem Assessment Survey (HICEAS). The original data was edited and annotated by the DCLDE 2022 working group (Yano et al. 2018) [https://www.soest.hawaii.edu/ore/dclde/dataset/]. We further processed the audio data into spectrograms suitable for deep learning applications.
We would like to express our gratitude to the contributors of the following codebases and data products: