TOMM

MKVSE: Multimodal Knowledge Enhanced Visual-semantic Embedding for Image-text Retrieval