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

Qingqing Cao, Mahyar Najibi, Sachin Mehta

Browse the full ICML paper archive.