metadata
tags:
- audio
Part of MONSTER: https://arxiv.org/abs/2502.15122.
CornellWhaleChalleng consists of hydrophone recordings [1]. The processed dataset consists of 30,000 (univariate) time series, each of length 4,000. The task is to distinguish right whale calls from other noises. (An abridged version of this dataset is included in the broader UCR archive.) This version of the dataset has been divided into stratified random cross-validation folds.
[1] André Karpištšenko, Eric Spalding, and Will Cukierski. (2013). The Marinexplore and Cornell University whale detection challenge. https://kaggle.com/competitions/whale-detection-challenge. Copyright 2011 Cornell University and the Cornell Research Foundation.