Arabic Sign Language

Dataset Overview

1 min read

Dataset Overview

data dataset karsl

The project utilizes the KArSL-502 (King Saud University Arabic Sign Language) dataset, a large-scale word-level sign language dataset.

Dataset Characteristics

  • Size: 502 classes (Arabic sign words).
  • Subjects: 3 distinct signers (01, 02, 03).
  • Structure:
    • Raw Video: Folder structure signer/split/word/video.mp4.
    • Keypoints: Extracted MediaPipe landmarks stored as .npz files.

Directory Structure

The processed dataset is expected to follow this hierarchy:

data/
├── 01/             # Signer 01
│   ├── train/
│   │   ├── 0001/   # Word ID 1
│   │   └── ...
│   └── test/
├── 02/             # Signer 02
└── 03/             # Signer 03

Keypoint Data Format

Each video is converted into a sequence of keypoints with the following shape: (SEQ_LEN, TOTAL_FEATURES)

Where TOTAL_FEATURES is the flattened vector of all landmarks:

  • Pose: 33 points × 3 coords (x, y, z)
  • Face: Selected subset points × 3 coords
  • Hands: 21 points × 3 coords × 2 hands

Graph View

Backlinks