Figure: Overview of ROMA's Streaming Dataset. This repository contains the Proactive subset (Green and Purple sections).
This repository contains the Proactive Interaction subset of the dataset introduced in the paper ROMA: Real-time Omni-Multimodal Assistant with Interactive Streaming Understanding.
This dataset is designed to train multimodal models for streaming video understanding, specifically focusing on tasks… See the full description on the dataset page:
https://huggingface.co/datasets/EurekaTian/ROMA_proactive.
