Video Descriptions with Temporal Annotations

The dataset connects 20,000 videos to temporally annotated sentence descriptions. On average, each video contains 3.65 temporally localized sentences describing unique segments and multiple events.

Use Cases

• Train a model to generate temporally localized sentence descriptions for video segments.
• Analyze the co-occurrence and duration of multiple events described within video captions.
• Develop video retrieval systems using the annotated sentence descriptions as queries.

Strengths

• Contains 20,000 videos.
• Each video has an average of 3.65 temporally annotated sentence descriptions.
• Descriptions cover unique segments and multiple events, including co-occurring ones.

Limitations

• The specific columns and data structure are unknown.
• The total number of annotated sentences is not provided.
• Geographic and temporal coverage details are unspecified.

Provenance

Source

HuggingFaceM4

Collection Method

Crowdsourced language creation with expert-generated annotations.

Freshness

Last updated on 2022-10-23.

Geography

Region is tagged as 'us'.