facebook/2M-Flores-ASL

2M-Flores数据集是2M-Belebele项目的一部分，包含了美国手语（ASL）的视频记录，这些视频是对原始flores200数据集中`dev`和`devtest`句子的手语翻译。数据集的创建过程包括将英语句子翻译成ASL、创建手语注释（glosses）、以及录制视频。视频在统一的背景下以高分辨率和高帧率录制，以确保质量。数据集的特征包括句子、手语注释、视频文件名等。

The 2M-Flores dataset is part of the 2M-Belebele project and contains ASL video recordings for all `dev` and `devtest` sentences in the original flores200 dataset. The dataset creation process involves providing English text to ASL translators and native signers, asking them to translate these sentences into ASL, create glosses, and record their interpretations. The glosses undergo additional quality checks by expert annotators to harmonize the glossing format. Videos are recorded against plain monochrome backgrounds with signers wearing monochrome upper body clothing to eliminate visual bias. All videos are captured in 1920x1080p resolution at 60 frames per second. The dataset columns include fields copied from the flores dataset as well as new fields such as gloss and signer.