1 paper · 1 filter
Jie Zhang, Qilang Ye, Hao Zhou +2
The dominant paradigm in video retrieval relies on embedding-based full-corpus scanning, which suffers from inherent computational inefficiency and the semantic asymmetry between i…