Moroccan Sign Language (MSL) dataset
Description
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.