CtrlSynth: Controllable Image Text Synthesis for Data-Efficient Multimodal Learning

CLIP
DiffM
Re-assign community