JingyunLiang / MANet

Official PyTorch code for Mutual Affine Network for Spatially Variant Kernel Estimation in Blind Image Super-Resolution (MANet, ICCV2021)

Home Page:https://arxiv.org/abs/2108.05302

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

some question about calculating psnr_kernel_y in stage1

leeziqiang opened this issue · comments

您好,请教个问题,在stage1阶段的validation和测试阶段,计算kernel的psnr_kernel_y 是选取了一个点的kernel来计算的,不用选取图像的所有的点的kernel吗,validation选的是visuals['KE'][5000, :, :],而test阶段是visuals['KE'][300, :, :],不是很理解,另外我的代码在stage1的时候在验证集上的psnr_kernel_y 不是很稳定,在53,54,55之间变化。

visuals['KE'][5000, :, :] is just for reference. We use LR image PSNR for evaluation. See Sec. 4.1 (performance evaluation) for more details.

ezoic increase your site revenue