KSL WORD BASED POSE DATASET

The KSL Pose Dataset is designed for computational analysis of Kenyan Sign Language (KSL) movements, leveraging pose estimation using MediaPipe. Each word in the dataset has a corresponding stickman video—generated from skeletal keypoints detected by MediaPipe—and NumPy files storing precise pose data, including joint coordinates, temporal trajectories, and kinematic descriptors. These pose representations are extracted frame-by-frame, capturing fine-grained motion patterns. The dataset enables spatiotemporal modeling for sign language recognition, facilitating pose-based feature extraction, sequence learning (e.g., LSTMs, Transformers), and multi-modal fusion techniques. It serves as a critical resource for computer vision, gesture analysis, and sign language processing research.