Fine-Grained Badminton Hit Instance Dataset

20,757 annotated hit instances provide a detailed breakdown of player actions within 2,066 rallies across 70 full-match videos. This subset of the FineBadminton collection was released by Moujuruo and updated in April 2026.

Use Cases

• Train models to classify hit instances (e.g., smash, drop shot) from video frames.
• Analyze rally patterns by sequencing annotated hit events across 2,066 rallies.
• Benchmark temporal action localization algorithms on 70 full-length match videos.

Strengths

• Contains 20,757 annotated hit instances for granular analysis.
• Includes 70 full-match videos providing complete game context.
• Annotations cover 2,066 rallies, enabling pattern analysis.

Limitations

• Subset size is 20K instances, smaller than the full 33K annotation collection.
• No column or feature details are provided, limiting understanding of annotation schema.
• Geographic and player diversity of the 70 source videos is unknown.

Provenance

Source

Subset of the FineBadminton dataset.

Collection Method

Annotations derived from full-match videos.

Freshness

Last updated April 2026.