Cross-modal Active Complementary Learning with Self-refining Correspondence | Read Paper on Bytez