Preprocessed skeleton sequences for Human Action Recognition. The dataset is derived from the HMDB51 video collection, focusing on pose estimation data for computer vision tasks.
Use Cases
- Train a temporal model on skeleton sequences for 51-class human action recognition.
- Benchmark pose estimation algorithms using standardized skeleton tracking data.
- Analyze action-specific kinematic patterns within preprocessed skeleton sequences.
Strengths
- Data is preprocessed, reducing initial feature engineering effort.
- Derived from the established HMDB51 benchmark video collection.
- Targeted for Skeleton Tracking and Human Action Recognition tasks.
Limitations
- The specific number of skeleton sequences, rows, or features is unknown.
- Dataset scope is limited to the 51 action categories defined in the original HMDB51 source.
- Lacks metadata on data collection method, author, or license details.
Provenance
- Source
- Derived from the HMDB51 video dataset.
- Collection Method
- Preprocessed from video sources, likely using pose estimation techniques.
- Time Range
- null
- Freshness
- null
- Geography
- null