36k SIBI 3D Hand Landmark Dataset for Fast Sign Language Recognition

This dataset contains 36,000+ rows of 3D hand landmark coordinates extracted using MediaPipe Hands for Sistem Isyarat Bahasa Indonesia (SIBI) alphabets (A–Z). Designed for real-time sign language translation, the dataset uses a Knowledge-Based Feature Extraction approach to extract 21 keypoints (x, y, z) per frame, normalized relative to the Wrist Joint for position-independent gesture classification (Translation Invariance). Benchmark evaluation using Deep Neural Network (DNN) achieved 100% accuracy on 6,919 test samples.