DIV2KRK

DIV2K Random Kernel

Introduced 2019-09-14

Using the validation set (100 images) from the widely used DIV2K dataset, we blurred and subsampled each image with a different, randomly generated kernel. Kernels were 11x11 anisotropic gaussians with random lengths λ1, λ2∼U(0.6, 5) independently distributed for each axis, rotated by a random angle θ∼U[−π, π].

Related Benchmarks

DIV2KRK - 2x upscaling/10-shot image generation/PSNR DIV2KRK - 2x upscaling/10-shot image generation/SSIM DIV2KRK - 2x upscaling/Image Reconstruction/PSNR DIV2KRK - 2x upscaling/Image Reconstruction/SSIM DIV2KRK - 2x upscaling/Image Restoration/PSNR DIV2KRK - 2x upscaling/Image Restoration/SSIM DIV2KRK - 4x upscaling/10-shot image generation/PSNR DIV2KRK - 4x upscaling/10-shot image generation/SSIM DIV2KRK - 4x upscaling/Image Reconstruction/PSNR DIV2KRK - 4x upscaling/Image Reconstruction/SSIM DIV2KRK - 4x upscaling/Image Restoration/PSNR DIV2KRK - 4x upscaling/Image Restoration/SSIM