Separating Chinese Character from Noisy Background Using GAN

Table 1

IoU of the separation results of printed characters by different deep learning methods. The IoU of printed and the IoU of background are first calculated for each result against ground truth in the test set and then averaged. The overall IoU is the average of the first two values.

IoU	CycleGAN	Pix2pix	U-Net	ESRGAN	DESRGAN
Printed	0.631	0.755	0.697	0.905	0.911
Background	0.869	0.935	0.910	0.977	0.978
Overall	0.750	0.845	0.803	0.941	0.944