基于生成对抗网络和坐标注意力机制的文本生成图像算法

李云红,张琪琪,陈锦妮,陈伟重,苏雪平,梁成名

Text-to-image generation algorithm based on generative adversarial network and coordinate attention mechanism

Yunhong LI,Qiqi ZHANG,Jinni CHEN,Weichong CHEN,Xueping SU,Chengming LIANG

表 3 CAT-GAN在Oxford-102、CUB-200和COCO数据集上的消融实验结果

Tab.3 Ablation experiment result of CAT-GAN on Oxford-102, CUB-200 and COCO dataset

方法 Oxford-102 CUB-200 COCO IS FID IS FID IS FID Baseline 3.31 24.32 3.37 23.71 25.12 34.11 Baseline+Ca 3.37 23.82 3.43 22.68 24.96 33.26 Baseline+CA，N = 1 3.42 23.67 3.65 22.34 25.16 33.01 Baseline+SRU，C = 1 3.38 23.65 3.46 22.59 25.10 32.89 Baseline+Ca+CA+SRU，N = 4，C = 3 3.53 20.06 4.48 18.42 26.45 29.36 Baseline+Ca+CA+SRU，N = 6，C = 5 3.79 17.69 4.97 16.17 26.96 27.13 Baseline+Ca+CA+SRU，N =8，C = 7 3.82 16.92 5.22 14.43 27.32 26.67 Baseline+Ca+CA+SRU，N = 10，C = 9 3.78 17.64 4.93 14.94 26.88 26.98