Indian Sign Language Mediapipe Holistic Landmarks for 4,284 Videos

Per-frame face, pose, and hand keypoints extracted from 4,284 Indian Sign Language videos. The dataset likely contains landmark coordinates generated by the Mediapipe Holistic model for each video frame. The author, organization, and specific collection details are unknown.

Use Cases

• Train sign language recognition models based on per-frame keypoint sequences.
• Benchmark pose estimation algorithms on sign language video data.
• Develop gesture classification systems using combined face, pose, and hand landmarks.
• Analyze the kinematics and articulation patterns of Indian Sign Language.

Strengths

• Contains keypoints for 4,284 distinct sign language videos.
• Provides multimodal landmarks for face, pose, and hands per frame, offering a holistic view of signing.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and data scale per video are unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Kaggle

Collection Method

Keypoints were likely extracted using the Mediapipe Holistic model from sign language videos.

Geography

India (implied by 'Indian Sign Language')