Clotho

Clotho是一个音频字幕数据集。数据集关注音频内容，以及字幕的多样性，由4,981个音频样本组成，每个音频样本5个字幕 (共24,905个字幕)，时长15 ~ 30秒，字幕长度8 ~ 20个字。

Clotho is an audio captioning dataset. Centered on audio content and the diversity of its captions, the dataset consists of 4,981 audio samples, with each sample paired with 5 captions, resulting in a total of 24,905 captions. Each audio clip has a duration ranging from 15 to 30 seconds, and each caption contains 8 to 20 Chinese characters.

搜集汇总

背景与挑战

背景概述

Clotho是一个音频字幕数据集，由坦佩雷大学于2019年发布，包含4,981个音频样本，每个样本对应5个字幕，共计24,905个字幕。音频时长在15至30秒之间，字幕长度范围为8到20个字。