Transactional Filipino Sign Language Dataset
Description
This dataset contains 1,065 .npy coordinates from video samples of Filipino Sign Language (FSL) gestures specifically curated for transactional and commercial interactions. The data was collected to support the development of Sign Language Recognition (SLR) systems aimed at bridging communication gaps in public facilities, markets, and government offices. The dataset consists of 26 distinct classes including numbers (1–10) and common transactional phrases: Yes, No, Thank You, Welcome, Again, How Much, How Many, Coin, Cash, Please, Wait, Card, Receipt, Discount, Hello, and Problem. This dataset is suitable for training and validating deep learning models (such as LSTM, GRU, or Transformers) for dynamic gesture recognition, particularly for edge computing applications in accessibility technology.
Files
Steps to reproduce
Data was captured in a controlled environment with consistent lighting using a Web Camera. Participants were oriented to perform the signs within a 1-meter distance from the camera. The raw video feeds were processed using MediaPipe Holistic to extract keypoints, which were then normalized and zero-padded to ensure a consistent input shape (100 frames) for neural network training. Data augmentation (flipping and scaling) was applied to select samples to increase diversity. These coordinates are taken as the content of the .npy files for each sample, which are the necessary files for model training.
Institutions
- Mapua University