PLM-Video Human: Human-Annotated Video Data for Vision Language Models

PLM-Video-Human is a collection of human-annotated resources for training Vision Language Models, focused on detailed video understanding. The dataset, created by Facebook, was last updated on May 21, -2025. Training tasks include fine-grained open-ended question answering, region-based video captioning, dense captioning, and temporal localization.

Use Cases

• Fine-tuning models for fine-grained open-ended question answering based on video content.
• Training region-based video captioning models based on annotated video segments.
• Developing models for region-based dense video captioning based on detailed scene descriptions.
• Conducting research on region-based temporal localization in videos.

Strengths

• Human-annotated resources, which likely provide high-quality training signals.
• Focuses on multiple complex video understanding tasks, suggesting a multifaceted design.
• Last updated on May 21, -2025, indicating recent maintenance.

Limitations

• Dataset structure, column definitions, and sample data are unavailable for inspection.
• Row count, file formats, and license information are unknown.
• The description is incomplete, requiring users to visit the external page for full details.

Provenance

Source

facebook

Collection Method

Human-annotated collection.

Time Range

null

Freshness

Last updated 2025-05-21 22:54:41

Geography

null