Lip Reading in the Wild (LRW)

数据集由多达1000个不同单词的话语组成，由数百个不同的说话者说出。所有视频的长度均为29帧 (1.16秒)，并且单词出现在视频的中间。单词持续时间在元数据中给出，您可以从中确定开始帧和结束帧。下表给出了数据集统计数据。

This dataset comprises utterances with up to 1000 distinct words, uttered by hundreds of different speakers. All videos have a fixed duration of 29 frames (1.16 seconds), and the target word appears at the center of each video. The duration of the target word is provided in the metadata, allowing the determination of its start and end frames. The statistical summary of this dataset is presented in the following table.

搜集汇总

背景与挑战

背景概述

Lip Reading in the Wild (LRW)是由牛津大学于2016年发布的数据集，包含1000个不同单词的话语，由数百个说话者说出，视频长度统一为29帧（1.16秒），单词出现在视频中间，持续时间在元数据中提供。