COSA: Concatenated Sample Pretrained Vision-Language Foundation Model
This addresses the problem of limited video-text data for researchers and practitioners in multimodal AI, though it is an incremental improvement using existing image-text corpora.
The paper tackles the lack of temporal modeling in vision-language foundation models by proposing COSA, which concatenates image-text pairs to create pseudo video-paragraph data, achieving state-of-the-art results on tasks like retrieval and captioning.
Due to the limited scale and quality of video-text training corpus, most vision-language foundation models employ image-text datasets for pretraining and primarily focus on modeling visually semantic representations while disregarding temporal semantic representations and correlations. To address this issue, we propose COSA, a COncatenated SAmple pretrained vision-language foundation model. COSA jointly models visual contents and event-level temporal cues using only image-text corpora. We achieve this by sequentially concatenating multiple image-text pairs as inputs for pretraining. This transformation effectively converts existing image-text corpora into a pseudo long-form video-paragraph corpus, enabling richer scene transformations and explicit event-description correspondence. Extensive experiments demonstrate that COSA consistently improves performance across a broad range of downstream tasks, including long-form/short-form video-text tasks and image-text tasks such as retrieval, captioning, and question answering. Notably, COSA achieves state-of-the-art results on various competitive benchmarks. Code and model are released at https://github.com/TXH-mercury/COSA.