VOYA_VSL: Vietnamese Sign Language Keypoint Sequences for Recognition

Approximately 161 samples of Vietnamese Sign Language videos, preprocessed into sequences of keypoints. The dataset was created by Kateht and last updated on 2025-09-23. Each sample is a matrix of 60 frames with 1605 keypoints or features per frame, stored in a NumPy compressed (.npz) format.

Use Cases

• Training sign language recognition models based on sequences of keypoints.
• Developing gesture classification algorithms based on the 60-frame video sequences.
• Benchmarking computer vision models for action recognition in sign language.
• Researching feature extraction and representation for sign language using the 1605 features per frame.

Strengths

• Contains approximately 161 preprocessed samples for sign language research.
• Data is structured as sequences of 60 frames with 1605 keypoints/features per frame, providing a consistent format.
• Last updated on 2025-09-23, indicating recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset size of ~161 samples may be limited for training complex deep learning models.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface

Collection Method

Collected and preprocessed for research in sign language recognition, with keypoints extracted using MediaPipe.

Freshness

Last updated 2025-09-23 02:50:54; freshness should be verified.

Geography

Vietnam (inferred from Vietnamese Sign Language focus)