HyperAI超神经

Image Super Resolution

图像超分辨率是一种机器学习任务,旨在将低分辨率图像放大至高分辨率,通常放大倍数为4倍或更高,同时尽可能保持图像的内容和细节。该技术能够显著提升图像质量,增强视觉效果,并提高计算机视觉算法的准确性,适用于多种应用场景。

2x upscaling
3x upscaling
4x upscaling
B100 - 2x upscaling
B100 - 3x upscaling
ML-CrAIST
B100 - 4x upscaling
BSD100 - 16x upscaling
ABPN
BSD100 - 2x upscaling
WaveMixSR-V2
BSD100 - 3x upscaling
HAT
BSD100 - 4x upscaling
Config (e)
BSD100 - 8x upscaling
DRLN+
BSD200 - 2x upscaling
BSDS100 - 2x upscaling
DBPN-RES-MR64-3
BSDS100 - 4x upscaling
DBPN-RES-MR64-3
BSDS100 - 8x upscaling
DBPN-RES-MR64-3
Celeb-HQ 4x upscaling
Edge-informed SR
CelebA
CelebA-HQ 128x128
IDM
Chikusei Dataset
DIP-HyperKite (ours)
CUFED5 - 4x upscaling
DIV2K val - 16x upscaling
ABPN
DIV2K val - 4x upscaling
AESOP
DIV2K val - 8x upscaling
DIV8K test - 16x upscaling
RFB-ESRGAN
DIV8K val - 16x upscaling
Ours w/o cycle-loss
EPFL NIR-VIS
RAMS (ours)
FFHQ 1024 x 1024 - 4x upscaling
FFHQ 256 x 256 - 4x upscaling
HiFaceGAN
FFHQ 512 x 512 - 4x upscaling
HiFaceGAN
General-100 - 4x upscaling
General100 - 4x upscaling
SROOE
General100 - 8x upscaling
FxSR-PD t=0.8
ImageNet
DAVI
IXI
EDSR+MMHCA
KITTI 2012 - 2x upscaling
KITTI 2012 - 4x upscaling
PASSRnet
KITTI 2015 - 2x upscaling
KITTI 2015 - 4x upscaling
PASSRnet
Manga109 - 16x upscaling
ABPN
Manga109 - 2x upscaling
DRCT-L
Manga109 - 3x upscaling
HMA†
Manga109 - 4x upscaling
SwinIR
Manga109 - 8x upscaling
DBPN-RES-MR64-3
Middlebury - 2x upscaling
Middlebury - 4x upscaling
PASSRnet
PIRM-test
RankSRGAN
Set14
ATD
Set14 - 2x upscaling
DRCT-L
Set14 - 3x upscaling
DRLN+
Set14 - 4x upscaling
DRCT-L
Set14 - 8x upscaling
DBPN-RES-MR64-3
Set5 - 2x upscaling
HAT-L
Set5 - 3x upscaling
HMA†
Set5 - 4x upscaling
Set5 - 5x upscaling
Set5 - 6x upscaling
Set5 - 8x upscaling
DBPN-RES-MR64-3
ShipSpotting
StableShip
Sun80 - 4x upscaling
Extracter-rec
Urban100 - 16x upscaling
ABPN
Urban100 - 2x upscaling
Urban100 - 3x upscaling
DRLN+
Urban100 - 4x upscaling
HMA†
Urban100 - 8x upscaling
DRLN+
USR-248 - 4x upscaling
SRDRM-GAN
VggFace2 - 8x upscaling
Full-GWAInet
WebFace - 8x upscaling
GFRNet
WLFW
ArcFace (0.4)