Dataset · nvidia
Full-duplex video communication benchmark for AI
nvidia/video-full-duplex-benchmark
Annotated English video calls from NVIDIA for evaluating conversational multimodal agents.
What's inside
The benchmark contains two-person video conference recordings (fewer than 1,000 examples) featuring English audio, video, and text. Data underwent three-stage manual annotation and covers 11 types of verbal and non-verbal dialog dynamics.
What you can do with it
- Evaluating multimodal AI agents capable of real-time live dialogue
- Testing models for recognizing non-verbal cues and interruptions in video calls
- Tasks
- Audio → audio
- Languages
- English
- Size
- up to 1K rows
- Downloads this month
- 1K
- Likes
- 28
- Access
- on request on Hugging Face
How to load it
from datasets import load_dataset
ds = load_dataset("nvidia/video-full-duplex-benchmark")