VideoLLM-online: Online Video Large Language Model for Streaming Video

TL;DR

This paper introduces VideoLLM-online, a novel framework for real-time, long-context streaming video dialogue, enabling large language models to process continuous video streams efficiently and effectively.

Contribution

It proposes the LIVE framework with new training, data generation, and inference methods for streaming video dialogue, and demonstrates significant real-time processing capabilities.

Findings

Supports over 10 FPS on 5-minute videos

Achieves state-of-the-art results on offline benchmarks

Enables long-context, real-time video conversations

Abstract

Recent Large Language Models have been enhanced with vision capabilities, enabling them to comprehend images, videos, and interleaved vision-language content. However, the learning methods of these large multimodal models typically treat videos as predetermined clips, making them less effective and efficient at handling streaming video inputs. In this paper, we propose a novel Learning-In-Video-Stream (LIVE) framework, which enables temporally aligned, long-context, and real-time conversation within a continuous video stream. Our LIVE framework comprises comprehensive approaches to achieve video streaming dialogue, encompassing: (1) a training objective designed to perform language modeling for continuous streaming inputs, (2) a data generation scheme that converts offline temporal annotations into a streaming dialogue format, and (3) an optimized inference pipeline to speed up the model responses in real-world video streams. With our LIVE framework, we built VideoLLM-online model upon Llama-2/Llama-3 and demonstrate its significant advantages in processing streaming videos. For instance, on average, our model can support streaming dialogue in a 5-minute video clip at over 10 FPS on an A100 GPU. Moreover, it also showcases state-of-the-art performance on public offline video benchmarks, such as recognition, captioning, and forecasting. The code, model, data, and demo have been made available at https://showlab.github.io/videollm-online.