DIBS: Enhancing Dense Video Captioning with Unlabeled Videos via Pseudo Boundary Enrichment and Online Refinement
This work addresses the problem of generating accurate event captions and boundaries in videos for applications like video understanding, offering a more data-efficient solution compared to previous approaches.
The paper tackles dense video captioning by proposing DIBS, a pretraining framework that uses unlabeled videos to generate and refine pseudo event boundaries and captions, achieving state-of-the-art results on YouCook2 and ActivityNet datasets with only 0.4% of the unlabeled data used by prior methods.
We present Dive Into the BoundarieS (DIBS), a novel pretraining framework for dense video captioning (DVC), that elaborates on improving the quality of the generated event captions and their associated pseudo event boundaries from unlabeled videos. By leveraging the capabilities of diverse large language models (LLMs), we generate rich DVC-oriented caption candidates and optimize the corresponding pseudo boundaries under several meticulously designed objectives, considering diversity, event-centricity, temporal ordering, and coherence. Moreover, we further introduce a novel online boundary refinement strategy that iteratively improves the quality of pseudo boundaries during training. Comprehensive experiments have been conducted to examine the effectiveness of the proposed technique components. By leveraging a substantial amount of unlabeled video data, such as HowTo100M, we achieve a remarkable advancement on standard DVC datasets like YouCook2 and ActivityNet. We outperform the previous state-of-the-art Vid2Seq across a majority of metrics, achieving this with just 0.4% of the unlabeled video data used for pre-training by Vid2Seq.