PLM-Video Auto: Synthetic Video Captions and Multiple-Choice Questions

Synthetic annotations for video understanding tasks, covering the YT-1B and Ego4d datasets. The dataset includes video captions and multiple-choice question-answer pairs, as described in the associated technical report. It was created by Facebook and last updated on the Hugging Face platform in April 2025.

Use Cases

• Training video captioning models based on synthetic text descriptions.
• Benchmarking video question-answering models based on multiple-choice questions.
• Developing multimodal retrieval systems based on video and text identifiers.
• Pre-training large language models on video-text data based on synthetic annotations.

Strengths

• Covers two major video datasets: YT-1B and Ego4d.
• Provides multiple annotation types: captions and multiple-choice questions.
• Dataset was last updated on 2025-04-21, indicating recent maintenance.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Row count, file formats, and column-level documentation are absent.
• Data is synthetic, which may introduce biases or artifacts not present in human annotations.

Provenance

Source

Facebook

Collection Method

Synthetic generation, as referenced in the associated technical report.

Time Range

null

Freshness

Last updated 2025-04-21 17:31:11

Geography

null