VLCap: Vision-Language with Contrastive Learning for Coherent Video Paragraph Captioning
This addresses the challenge of video understanding and captioning for applications like video indexing and accessibility, but it appears incremental as it builds on existing vision-language approaches with contrastive learning.
The paper tackles the problem of generating coherent paragraph descriptions for untrimmed videos by leveraging vision-language interaction, and the result shows that their VLCap method outperforms existing state-of-the-art methods on accuracy and diversity metrics on ActivityNet Captions and YouCookII datasets.
In this paper, we leverage the human perceiving process, that involves vision and language interaction, to generate a coherent paragraph description of untrimmed videos. We propose vision-language (VL) features consisting of two modalities, i.e., (i) vision modality to capture global visual content of the entire scene and (ii) language modality to extract scene elements description of both human and non-human objects (e.g. animals, vehicles, etc), visual and non-visual elements (e.g. relations, activities, etc). Furthermore, we propose to train our proposed VLCap under a contrastive learning VL loss. The experiments and ablation studies on ActivityNet Captions and YouCookII datasets show that our VLCap outperforms existing SOTA methods on both accuracy and diversity metrics.