Exploring multi-semantic disentangled controls in GANs using conjugate gradient optimization

Zhenbo Yu, Jian Jin, Zhenyong Fu, Jian Yang

Abstract: Highlights•Propose QCGO to better disentangle different semantics in the latent space of GANs.•QCGO can accelerate convergence speed to achieve rapid image editing.•QCGO can be employed to fulfill real image manipulation.•QCGO’s effectiveness has been validated on diverse models and datasets.•Experimental results show that QCGO outperforms the SOTA.