GroundVTS: Visual Token Sampling in Multimodal Large Language Models for Video Temporal Grounding
This work improves video understanding for applications like video analysis and retrieval, but it is incremental as it builds on existing Vid-LLM architectures with a novel sampling method.
The paper tackled the problem of video temporal grounding in multimodal large language models by addressing the loss of crucial temporal cues due to uniform frame sampling, resulting in a 7.7-point improvement in mIoU for moment retrieval and a 12.0-point improvement in mAP for highlight detection on standard benchmarks.
Video temporal grounding (VTG) is a critical task in video understanding and a key capability for extending video large language models (Vid-LLMs) to broader applications. However, existing Vid-LLMs rely on uniform frame sampling to extract video information, resulting in a sparse distribution of key frames and the loss of crucial temporal cues. To address this limitation, we propose Grounded Visual Token Sampling (GroundVTS), a Vid-LLM architecture that focuses on the most informative temporal segments. GroundVTS employs a fine-grained, query-guided mechanism to filter visual tokens before feeding them into the LLM, thereby preserving essential spatio-temporal information and maintaining temporal coherence. Futhermore, we introduce a progressive optimization strategy that enables the LLM to effectively adapt to the non-uniform distribution of visual features, enhancing its ability to model temporal dependencies and achieve precise video localization. We comprehensively evaluate GroundVTS on three standard VTG benchmarks, where it outperforms existing methods, achieving a 7.7-point improvement in mIoU for moment retrieval and 12.0-point improvement in mAP for highlight detection. Code is available at https://github.com/Florence365/GroundVTS.