Skip to main content
. 2019 Aug 22;10(9):4742–4756. doi: 10.1364/BOE.10.004742

Fig. 4.

Fig. 4.

Training procedures of kmax evaluation and fine-tuning. (A) Generator loss curve with and without TG loss constraint. Different kmax values (1, N/2, and N) was set for the TG loss in the first training procedure. (B) PSNR (dB) of the validation dataset with and without TG loss constraint. (C) Generator loss curve in the fine-tuning procedure. kmax was set to be N/2 for the TG loss. (D) PSNR (dB) of the validation dataset in the fine-tuning procedure.