Multichannel Cross-Scale Semantic Coherent Attention Network for Image Inpainting

Table 2

Comparison result of different methods.

Index	Dataset	Urban100			DTD			CelebA
	Mask ratio	0.3	0.4	0.5	0.3	0.4	0.5	0.3	0.4	0.5
PSNR↑	RDN	25.8047	26.0044	27.5418	24.8502	26.1248	28.5307	25.3175	27.4980	29.2314
	DeepFill	24.9095	24.1774	24.8977	22.9332	25.4656	25.6100	28.4158	29.3279	28.8496
	Pconv	20.5415	20.1884	20.2059	21.1439	21.1842	20.5190	24.2375	24.0591	24.8409
	RFR	34.0451	37.3325	33.4948	29.9784	28.3995	30.6775	35.5869	32.5108	29.2596
	PSR	34.2673	36.6664	36.5071	35.8867	35.2267	34.9433	37.7040	33.6886	34.2690
	Our	34.6323	39.3942	37.8651	37.6460	36.2432	35.9936	38.6528	37.4934	36.6306
SSIM↑	RDN	0.8207	0.8203	0.8689	0.7442	0.8024	0.8571	0.8251	0.8752	0.9086
	DeepFill	0.7659	0.7274	0.7610	0.6847	0.7490	0.7344	0.8815	0.8982	0.8639
	Pconv	0.6405	0.6081	0.6229	0.6126	0.6365	0.6845	0.7659	0.7585	0.8067
	RFR	0.9747	0.9788	0.9512	0.9259	0.9092	0.9513	0.9288	0.9570	0.7723
	PSR	0.9798	0.9884	0.9834	0.9767	0.9746	0.9756	0.9845	0.9713	0.9779
	Our	0.9856	0.9883	0.9846	0.9794	0.9777	0.9773	0.9851	0.9813	0.9797
L1↓	RDN	0.0261	0.0316	0.0233	0.0379	0.0308	0.0234	0.0259	0.0195	0.0141
	DeepFill	0.0412	0.0494	0.0480	0.0520	0.0505	0.0431	0.0274	0.0250	0.0231
	Pconv	0.0743	0.0847	0.0733	0.0769	0.0722	0.0812	0.0480	0.0470	0.0453
	RFR	0.0178	0.0177	0.0192	0.0173	0.0148	0.0159	0.0182	0.0145	0.0131
	PSR	0.0163	0.0148	0.0133	0.0125	0.0116	0.0161	0.0107	0.0119	0.0134
	Our	0.0135	0.0096	0.0109	0.0100	0.0114	0.0107	0.0127	0.0132	0.0116