ProTA: Probabilistic Token Aggregation for Text-Video Retrieval

arXiv:2404.12216v1 Announce Type: new
Abstract: Text-video retrieval aims to find the most relevant cross-modal samples for a given query. Recent methods focus on modeling the whole spatial-temporal relations. However, since video clips contain more diverse content than captions, the model aligning these asymmetric video-text pairs has a high risk of retrieving many false positive results. In this paper, we propose Probabilistic Token Aggregation (\textit{ProTA}) to handle cross-modal interaction with content asymmetry. Specifically, we propose dual partial-related aggregation to disentangle and …

abstract aggregation arxiv captions cs.cv diverse false focus however modal modeling query relations retrieval risk samples spatial temporal text token type video