1 paper · 2 filters
Jingfeng Yang, Ziyang Wu, Yue Zhao +1
Currently, the most dominant approach to establishing language-image alignment is to pre-train text and image encoders jointly through contrastive learning, such as CLIP and its va…