Yoga Pose Keypoints Extracted via MediaPipe

Keypoints extracted from yoga pose videos using the MediaPipe framework. The dataset contains 82 keypoints per frame, saved in NumPy array format. The author, source organization, and total volume of data are unspecified.

Use Cases

• Train a pose classifier to identify yoga asanas using sequences of 82 keypoints per frame.
• Develop a keypoint-based anomaly detection model to flag incorrect yoga postures by analyzing joint angle deviations.
• Benchmark the accuracy of MediaPipe's pose estimation on complex, contorted yoga poses against ground truth data.
• Create a generative model for synthesizing realistic yoga pose sequences from the 82-dimensional keypoint time series.

Strengths

• Keypoints are extracted using MediaPipe, a widely-used and validated pose estimation library.
• Data is structured as 82 keypoints per observation, providing a detailed skeletal representation.

Limitations

• The total number of video samples, frames, and subjects is unknown, making scale assessment impossible.
• Potential label noise or estimation errors inherent to the MediaPipe pipeline are not quantified.
• Lack of demographic or contextual metadata limits analysis of pose variation across populations.

Provenance

Source

null

Collection Method

Keypoints extracted using MediaPipe from source videos.

Time Range

null

Freshness

null

Geography

null