VideoFDB: Annotated Two-Person Video Conference Recordings

VideoFDB is a benchmark dataset of annotated, two-person video conference recordings designed to support the evaluation of multimodal AI agents in conversational settings. The dataset covers 11 distinct conversational dynamics spanning verbal, nonverbal, and mixed-modality behavior, annotated through a three-pass human-in-the-loop pipeline. It was created by NVIDIA and last updated on HuggingFace in May 2026.

Use Cases

• Evaluating multimodal AI agents based on annotated conversational dynamics.
• Training models to recognize verbal and nonverbal cues in dyadic conversations.
• Benchmarking performance on mixed-modality behavior understanding in video calls.

Strengths

• Annotations cover 11 distinct conversational dynamics.
• Uses a three-pass human-in-the-loop annotation pipeline.
• Specifically designed for evaluating multimodal AI agents.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset consists of trimmed recordings; the original context may be missing.

Provenance

Source

NVIDIA

Collection Method

Annotated video conference recordings processed through a human-in-the-loop pipeline.

Freshness

Last updated 2026-05-29 15:22:22.