Improving Video Retrieval Using Multilingual Knowledge Transfer

Video retrieval has seen tremendous progress with the development of vision-language models. However, further improving these models require additional labelled data which is a huge manual effort. In this paper, we propose framework MKTVR, that utilizes knowledge transfer from multilingual model to boost performance video retrieval. We first use state-of-the-art machine translation construct pseudo ground-truth video-text pairs. then learn representation where English and non-English text queries are represented in common embedding space based on pretrained evaluate our proposed approach four datasets such as MSRVTT, MSVD, DiDeMo Charades. Experimental results demonstrate achieves all outperforming previous Finally, also video-retrieval dataset encompassing six languages show outperforms zero-shot setting.