Part of MONSTER: https://arxiv.org/abs/2502.15122.

CornellWhaleChallenge
Category	EEG
Num. Examples	12,289
Num. Channels	14
Length	256
Sampling Freq.	128 Hz
Num. Classes	2
License	Other
Citations	[1] [2]

CrowdSourced consists of EEG data collected as part of a study investigating brain activity during a resting state task, which included two conditions: eyes open and eyes closed, each lasting 2 minutes. The dataset contains EEG recordings from 60 participants, but only 13 successfully completed both conditions. The recordings were captured using 14-channel EEG headsets—specifically the Emotiv EPOC+, EPOC X, and EPOC devices. These devices provide high-quality, wireless brainwave data that is ideal for analyzing resting-state brain activity [1].

The data was initially recorded at a high frequency of 2048 Hz and later downsampled to 128 Hz for processing. To segment the data for analysis, we used a 2-second window (equivalent to 256 time steps) with a 32 time-step stride to capture the dynamics of brain activity while maintaining a manageable data size. The raw EEG data for the 13 participants, along with preprocessing steps, analysis scripts, and visualization tools, are openly available on the Open Science Framework [2]. The processed dataset consists of 12,289 multivariate time series, each of length 256 (i.e., representing 2 seconds of data per time series at a sampling rate of 128 Hz). This version of the dataset has been split into cross-validation folds based on participant.

[1] Nikolas S Williams, William King, Geoffrey Mackellar, Roshini Randeniya, Alicia McCormick, and Nicholas A Badcock. (2023). Crowdsourced EEG experiments: A proof of concept for remote EEG acquisition using emotivpro builder and emotivlabs. Heliyon, 9(8).

[2] Nikolas Scott Williams, William King, Roshini Randeniya, and Nicholas A Badcock. (2022). Crowdsourced. https://osf.io/9bvgh/.