CLIP-ViP: Adapting Pre-trained Image-Text Model to Video-Language Representation Alignment

TL;DR

This paper introduces CLIP-ViP, a novel method that adapts pre-trained CLIP for video-language tasks by addressing domain gap and data scale issues, achieving state-of-the-art results on multiple datasets.

Contribution

Proposes CLIP-ViP, a new approach for effective post-pretraining adaptation of CLIP to video-language tasks, with a Video Proxy mechanism and cross-modal learning.

Findings

Significant improvement in video-text retrieval performance.

Achieves SOTA results on MSR-VTT, DiDeMo, LSMDC, and ActivityNet.

Highlights the importance of data scale and domain gap mitigation.

Abstract

The pre-trained image-text models, like CLIP, have demonstrated the strong power of vision-language representation learned from a large scale of web-collected image-text data. In light of the well-learned visual features, some existing works transfer image representation to video domain and achieve good results. However, how to utilize image-language pre-trained model (e.g., CLIP) for video-language pre-training (post-pretraining) is still under explored. In this paper, we investigate two questions: 1) what are the factors hindering post-pretraining CLIP to further improve the performance on video-language tasks? and 2) how to mitigate the impact of these factors? Through a series of comparative experiments and analyses, we find that the data scale and domain gap between language sources have great impacts. Motivated by these, we propose a Omnisource Cross-modal Learning method equipped with a Video Proxy mechanism on the basis of CLIP, namely CLIP-ViP. Extensive results show that our approach improves the performance of CLIP on video-text retrieval by a large margin. Our model also achieves SOTA results on a variety of datasets, including MSR-VTT, DiDeMo, LSMDC, and ActivityNet. We will release our code and pre-trained CLIP-ViP models at https://github.com/microsoft/XPretrain/tree/main/CLIP-ViP.