WLASL

该数据集名为WLASL，是从网络视频中收集的美国手语数据集，总计包含2,000个词汇和21,083个样本，这些样本分别用于训练、验证和测试。此外，还发布了最频繁使用的前100个和前300个词汇，分别命名为WLASL100和WLASL300。该数据集的规模为2,000个词汇和21,083个样本，其任务重点是手语生成。

The dataset named WLASL is an American Sign Language (ASL) dataset collected from web videos. It contains a total of 2,000 vocabulary terms and 21,083 samples, which are split into training, validation and test sets respectively. Additionally, two subsets of the top 100 and top 300 most frequently used vocabulary terms, named WLASL100 and WLASL300 respectively, have also been released. With a scale of 2,000 vocabulary terms and 21,083 samples, this dataset focuses on sign language generation tasks.

搜集汇总

背景与挑战

背景概述

WLASL是一个大规模美国手语单词级别识别数据集，包含多个子集和丰富的视频样本，适用于手语识别和翻译研究。数据集提供了完整的下载、预处理和训练流程，并受到C-UDA许可的限制，仅限非商业用途。