SF2T: Self-supervised Fragment Finetuning of Video-LLMs for Fine-Grained Understanding
This addresses a key limitation in video-based AI for researchers and developers, though it appears incremental as it builds on existing Video-LLMs with a novel fine-tuning approach.
The paper tackles the problem of Video-LLMs struggling with fine-grained video understanding by proposing a self-supervised fragment fine-tuning method, which improves their ability to capture spatiotemporal details without labor-intensive annotations.
Video-based Large Language Models (Video-LLMs) have witnessed substantial advancements in recent years, propelled by the advancement in multi-modal LLMs. Although these models have demonstrated proficiency in providing the overall description of videos, they struggle with fine-grained understanding, particularly in aspects such as visual dynamics and video details inquiries. To tackle these shortcomings, we find that fine-tuning Video-LLMs on self-supervised fragment tasks, greatly improve their fine-grained video understanding abilities. Hence we propose two key contributions:(1) Self-Supervised Fragment Fine-Tuning (SF$^2$T), a novel effortless fine-tuning method, employs the rich inherent characteristics of videos for training, while unlocking more fine-grained understanding ability of Video-LLMs. Moreover, it relieves researchers from labor-intensive annotations and smartly circumvents the limitations of natural language, which often fails to capture the complex spatiotemporal variations in videos; (2) A novel benchmark dataset, namely FineVidBench, for rigorously assessing Video-LLMs' performance at both the scene and fragment levels, offering a comprehensive evaluation of their capabilities. We assessed multiple models and validated the effectiveness of SF$^2$T on them. Experimental results reveal that our approach improves their ability to capture and interpret spatiotemporal details.