bdanko/how2sign-landmarks-front-raw-parquet

该数据集是How2Sign项目的一部分，包含手语视频的RGB前端剪辑和相应的MediaPipe地标数据。具体来说，数据集包括两个部分：bdanko/how2sign-landmarks-front-raw-parquet提供原始的地标特征（如姿势、面部和手部关键点），以Parquet格式存储；bdanko/how2sign-rgb-front-clips提供对应的RGB视频剪辑，以MP4格式存储。这些数据用于手语识别和处理任务，支持从视频中提取和可视化地标信息。

This dataset is part of the How2Sign project, containing RGB front clips of sign language videos and corresponding MediaPipe landmark data. Specifically, the dataset consists of two components: bdanko/how2sign-landmarks-front-raw-parquet provides raw landmark features (such as pose, face, and hand keypoints) stored in Parquet format; bdanko/how2sign-rgb-front-clips provides corresponding RGB video clips stored in MP4 format. These data are used for sign language recognition and processing tasks, supporting the extraction and visualization of landmark information from videos.