real_input_mean_gray = real_input_img.float().mean() syn_lable_mean_gray = label_img.float().mean() real_img_adjust = torch.clamp((real_input_img * (syn_lable_mean_gray / real_input_mean_gray)), 0, 1)
