ChatVTG: Video Temporal Grounding via Chat with Video Dialogue Large Language Models

TL;DR

ChatVTG introduces a zero-shot video temporal grounding method using Video Dialogue LLMs to generate and match captions with queries, reducing reliance on annotated data and improving accuracy.

Contribution

This work pioneers the use of Video Dialogue LLMs for zero-shot VTG, combining coarse caption matching with moment refinement for precise results.

Findings

Outperforms existing zero-shot VTG methods on multiple datasets

Effectively generates multi-granularity segment captions without paired annotations

Achieves superior temporal grounding accuracy

Abstract

Video Temporal Grounding (VTG) aims to ground specific segments within an untrimmed video corresponding to the given natural language query. Existing VTG methods largely depend on supervised learning and extensive annotated data, which is labor-intensive and prone to human biases. To address these challenges, we present ChatVTG, a novel approach that utilizes Video Dialogue Large Language Models (LLMs) for zero-shot video temporal grounding. Our ChatVTG leverages Video Dialogue LLMs to generate multi-granularity segment captions and matches these captions with the given query for coarse temporal grounding, circumventing the need for paired annotation data. Furthermore, to obtain more precise temporal grounding results, we employ moment refinement for fine-grained caption proposals. Extensive experiments on three mainstream VTG datasets, including Charades-STA, ActivityNet-Captions, and TACoS, demonstrate the effectiveness of ChatVTG. Our ChatVTG surpasses the performance of current zero-shot methods.