video-SALMONN: Speech-Enhanced Audio-Visual Large Language Models

TL;DR

video-SALMONN is an end-to-end audio-visual large language model that effectively understands speech, visual, and audio elements in videos, significantly improving video question-answering accuracy and reasoning capabilities.

Contribution

The paper introduces a novel multi-resolution causal Q-Former and training schemes for integrated speech, audio, and visual understanding in av-LLMs, advancing video comprehension.

Findings

Achieves over 25% accuracy improvement on video-QA tasks.

Surpasses 30% accuracy on audio-visual QA with speech.

Demonstrates advanced reasoning abilities in video understanding.

Abstract

Speech understanding as an element of the more generic video understanding using audio-visual large language models (av-LLMs) is a crucial yet understudied aspect. This paper proposes video-SALMONN, a single end-to-end av-LLM for video processing, which can understand not only visual frame sequences, audio events and music, but speech as well. To obtain fine-grained temporal information required by speech understanding, while keeping efficient for other video elements, this paper proposes a novel multi-resolution causal Q-Former (MRC Q-Former) structure to connect pre-trained audio-visual encoders and the backbone large language model. Moreover, dedicated training approaches including the diversity loss and the unpaired audio-visual mixed training scheme are proposed to avoid frames or modality dominance. On the introduced speech-audio-visual evaluation benchmark, video-SALMONN achieves more than 25\% absolute accuracy improvements on the video-QA task and over 30\% absolute accuracy improvements on audio-visual QA tasks with human speech. In addition, video-SALMONN demonstrates remarkable video comprehension and reasoning abilities on tasks that are unprecedented by other av-LLMs. Our training code and model checkpoints are available at \texttt{\url{https://github.com/bytedance/SALMONN/}}.