EPFL-Smart-Kitchen: Action Recognition Benchmark

The EPFL-Smart-Kitchen-30 dataset provides a benchmark for action recognition. It likely contains video clips of kitchen activities annotated with action labels. The dataset was created by amathislab and was last updated on October 23, 2025.

Use Cases

• Benchmarking action recognition models based on video clips.
• Comparing the effectiveness of 3D body pose, hand pose, and eye gaze modalities for recognition.
• Training models to classify specific kitchen-related human actions.

Strengths

• Based on the EPFL-Smart-Kitchen-30 dataset, which suggests a consistent underlying data source.
• Provides a recognition-oriented setup for comparing multiple input modalities.
• Last updated on October 23, 2025, indicating recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The full description is hosted externally, requiring an extra step for complete details.

Provenance

Source

amathislab

Freshness

Last updated 2025-10-23 09:24:49