MindBigData 2023 MNIST-8B

MindBigData 2023 MNIST-8B是由MindBigData创建的迄今为止最大的脑信号开放数据集，专为机器学习设计。该数据集基于单个受试者的EEG信号，使用128通道自定义设备捕捉，复现了Yaan LeCun等人的MNIST数据集中的70,000个数字。数据收集过程中，受试者观看原始数字像素并同时听到0到9的语音标签。数据集不仅包含EEG数据，还包括音频文件和电极的3D坐标，适用于多种研究目的，如标签分类、数字像素重建和基于脑信号的音频波形分析。此数据集的发布旨在增加对大脑的理解，并为机器学习社区提供新的挑战性指标，以促进新算法的开发和人工智能领域的进步。

MindBigData 2023 MNIST-8B is the largest open-source brain signal dataset created by MindBigData to date, specifically tailored for machine learning research. This dataset is based on electroencephalogram (EEG) signals from a single participant, captured using a custom 128-channel device, and reproduces the 70,000 digits from the MNIST dataset originally curated by Yann LeCun et al. During data collection, the participant viewed raw digit pixel stimuli while simultaneously listening to spoken labels corresponding to numbers 0 through 9. The dataset includes not only raw EEG data, but also audio files and 3D coordinates of the EEG electrodes, supporting a wide range of research applications such as label classification, digit pixel reconstruction, and audio waveform analysis based on brain signals. The release of this dataset aims to advance fundamental understanding of the human brain, provide the global machine learning community with novel challenging benchmarks, and foster the development of new algorithms and progress in the field of artificial intelligence.