基于条件生成对抗和泊松融合的图像数据增强方法

陈可嘉,曾威文,王朕卿

Image data enhancement method based on conditional generative adversarial network and Poisson blending

Kejia CHEN,Weiwen ZENG,Zhenqing WANG

表 1 不同生成网络的参数量和FLOPs对比

Tab.1 Comparison of parameters and FLOPs for different generative networks

生成网络 NpG/106 NpD/106 NpT/106 NF/109 PB-WCGAN-GP 3.66 2.63 6.31 0.72 StyleGAN 5.31 15.34 20.65 5.98 ACGAN 3.45 2.72 6.17 0.71 WGAN-Div 3.78 2.91 6.79 0.73 DDPM — — 34.12 8.91 R3GAN 11.58 11.51 23.09 6.19