SpokeN-100

SpokeN-100是由德国埃尔朗根的人工智能生物医学工程系创建的一个跨语言基准数据集，专注于不同语言中口语数字的分类。该数据集包含32位不同说话者用英语、普通话、德语和法语四种语言说出的从0到99的数字，总计12,800个音频样本。数据集通过先进的AI模型完全人工生成，确保了数据的一致性和质量。SpokeN-100特别适用于微型深度学习领域的算法评估和优化，尤其是在资源受限的设备上执行的紧凑深度学习模型。SpokeN-100 is a cross-lingual benchmark dataset created by the Department of Artificial Intelligence and Biomedical Engineering of Erlangen, Germany, focusing on spoken digit classification across different languages. This dataset contains 12,800 audio samples in total, with digits from 0 to 99 spoken by 32 distinct speakers in four languages: English, Mandarin, German, and French. The dataset is fully artificially generated using advanced AI models, ensuring data consistency and quality. SpokeN-100 is particularly suitable for algorithm evaluation and optimization in the field of tiny deep learning, especially for compact deep learning models deployed on resource-constrained devices.

搜集汇总

背景与挑战

背景概述

SpokeN-100是一个跨语言口语数字分类基准数据集，由德国埃尔朗根的人工智能生物医学工程系创建，包含英语、普通话、德语和法语四种语言的12,800个AI生成音频样本，覆盖0到99的数字。该数据集专为微型深度学习设计，适用于资源受限设备上的算法评估和优化，确保数据一致性和高质量。