Moroccan Sign Language (MSL) dataset

The dataset consists of annotated video clips of Moroccan Sign Language (MSL) collected from publicly available online videos. The videos were selected based on their clarity, inclusion of Arabic subtitles, and diversity in signers, encompassing both adults and children. The dataset includes 2,310 videos with a total duration of 7 hours, 15 minutes, and 19 seconds. It covers 2069 unique words and sentences. Each video is labeled with the corresponding Arabic subtitles and segmented into clips containing single words or phrases to ensure precise annotations. The dataset contains raw video files in .mp4 format, along with extracted keypoint data in .npy. These files include normalized 3D coordinates for 75 keypoints, covering the face, hands, and upper body. This dataset is tailored to support machine learning research in sign language recognition and gesture-based human-computer interaction.本数据集包含经标注的摩洛哥手语（Moroccan Sign Language, MSL）视频片段，素材均采集自公开可获取的在线视频。视频筛选标准涵盖画面清晰度、阿拉伯语字幕配置以及手语者多样性，覆盖成人与儿童两类群体。本数据集共包含2310段视频，总时长为7小时15分19秒，涵盖2069个独特词汇与语句。每段视频均配有对应的阿拉伯语字幕，且被切割为仅包含单个词汇或短语的片段，以保障标注精度。本数据集包含.mp4格式的原始视频文件，以及.npy格式的提取关键点数据。该关键点数据包含75个关键点的归一化三维坐标，覆盖面部、手部与上半身区域。本数据集专为支持手语识别与基于手势的人机交互领域的机器学习研究而构建。