
摘要
超分辨率生成对抗网络(SRGAN)是一项开创性的工作,能够在单图像超分辨率过程中生成逼真的纹理。然而,生成的细节往往伴随着令人不悦的伪影。为了进一步提升视觉质量,我们对SRGAN的三个关键组件——网络架构、对抗损失和感知损失进行了深入研究,并对每个组件进行了改进,从而提出了增强型SRGAN(ESRGAN)。具体而言,我们将无批量归一化的残差密集块中的残差引入作为基本网络构建单元(Residual-in-Residual Dense Block, RRDB)。此外,我们借鉴了相对论生成对抗网络的思想,使判别器预测相对真实性而非绝对值。最后,我们通过使用激活前的特征来改进感知损失,这可以为亮度一致性和纹理恢复提供更强的监督。得益于这些改进,所提出的ESRGAN在视觉质量上比SRGAN更加出色,生成的纹理更为逼真和自然,并在PIRM2018-SR挑战赛中获得了第一名。代码可在https://github.com/xinntao/ESRGAN 获取。
代码仓库
xinntao/ESRGAN
官方
pytorch
GitHub 中提及
Pulkitdzrt/ML-Image-Super-Resolution
GitHub 中提及
HyeongJu916/Boaz-SR-ESRGAN-PyTorch
pytorch
GitHub 中提及
Dictator890/ESRGAN-Android-Demo
GitHub 中提及
RezisEwig/esrgan
tf
GitHub 中提及
olaviinha/NeuralImageSuperResolution
GitHub 中提及
BearNinja123/esrGAN_vBearNinja123
tf
GitHub 中提及
aba450/Super-Resolution
pytorch
GitHub 中提及
kingcong/gpu_ESRGAN
mindspore
GitHub 中提及
sberbank-ai/real-esrgan
pytorch
GitHub 中提及
YanyingZH/srgan-esrgan-pytorch
pytorch
GitHub 中提及
yangyucheng000/ESRGAN
mindspore
GitHub 中提及
YvanG/VQGAN-CLIP
pytorch
GitHub 中提及
cviaai/PSF-GAN-ESTIMATION
pytorch
GitHub 中提及
u7javed/Image-Super-Resolution-Enhancer-via-ESRGAN
pytorch
GitHub 中提及
Gowti-AiboT/GAN-based-SISR-Keras
tf
GitHub 中提及
amogh7joshi/media-vision
pytorch
GitHub 中提及
FirstPotatoMan/TensorflowJS_ESRGAN
tf
GitHub 中提及
doantienthongbku/ESRGAN
pytorch
GitHub 中提及
AnjanaGJoseph/Super-Resolution-GAN
GitHub 中提及
sberbank-ai/DigiTeller
tf
GitHub 中提及
final-0/Real-ESRGAN
pytorch
peteryuX/esrgan-tf2
tf
GitHub 中提及
luciennnnnnn/dualformer
pytorch
GitHub 中提及
final-0/Real-ESRGAN-bicubic
pytorch
nabarunbaruaAIML/ATCC_Yolov5
pytorch
GitHub 中提及
eriklindernoren/PyTorch-GAN
pytorch
GitHub 中提及
sberbank-ai/DigiTales
tf
GitHub 中提及
CorentinMAG/SRGAN
GitHub 中提及
omkarghugarkar007/Neural_Super_Sampling
tf
GitHub 中提及
itsuki8914/ESRGAN-TensorFlow
tf
GitHub 中提及
idealo/image-super-resolution
tf
GitHub 中提及
Lornatang/ESRGAN-PyTorch
pytorch
GitHub 中提及
alililia/ascend_ESRGAN
mindspore
GitHub 中提及
fenghansen/ESRGAN-Keras
tf
GitHub 中提及
allenai/satlas-super-resolution
pytorch
GitHub 中提及
lizhuoq/ESRGAN-pytorch
pytorch
nannau/DoWnGAN
pytorch
GitHub 中提及
wwhio/megmodels
pytorch
sdauzcm/sr-basicsr
pytorch
GitHub 中提及
u7javed/Image-Enhancer-via-ESRGAN
pytorch
GitHub 中提及
基准测试
| 基准 | 方法 | 指标 |
|---|---|---|
| face-hallucination-on-ffhq-512-x-512-16x | ESRGAN | FID: 50.901 LPIPS: 0.3928 NIQE: 15.383 |
| image-super-resolution-on-bsd100-4x-upscaling | SRGAN + Residual-in-Residual Dense Block | PSNR: 27.85 SSIM: 0.7455 |
| image-super-resolution-on-ffhq-1024-x-1024-4x | ESRGAN | FID: 72.73 MS-SSIM: 0.782 PSNR: 19.84 SSIM: 0.353 |
| image-super-resolution-on-ffhq-256-x-256-4x | ESRGAN | FID: 166.36 MS-SSIM: 0.747 PSNR: 15.43 SSIM: 0.267 |
| image-super-resolution-on-ffhq-512-x-512-4x | ESRGAN | FED: 0.1107 FID: 3.503 LLE: 2.261 LPIPS: 0.1221 MS-SSIM: 0.935 NIQE: 6.984 PSNR: 27.134 SSIM: 0.741 |
| image-super-resolution-on-manga109-4x | bicubic | PSNR: 24.89 SSIM: 0.7866 |
| image-super-resolution-on-manga109-4x | SRGAN + Residual-in-Residual Dense Block | PSNR: 31.66 SSIM: 0.9196 |
| image-super-resolution-on-pirm-test | ESRGAN | NIQE: 2.55 |
| image-super-resolution-on-set14-4x-upscaling | SRGAN + Residual-in-Residual Dense Block | PSNR: 28.99 SSIM: 0.7917 |
| image-super-resolution-on-urban100-4x | SRGAN + Residual-in-Residual Dense Block | PSNR: 27.03 SSIM: 0.8153 |
| image-super-resolution-on-urban100-4x | bicubic | PSNR: 23.14 SSIM: 0.6577 |
| video-super-resolution-on-msu-video-upscalers | ESRGAN | PSNR: 27.29 SSIM: 0.936 VMAF: 56.69 |
| video-super-resolution-on-msu-vsr-benchmark | ESRGAN | 1 - LPIPS: 0.948 ERQAv1.0: 0.735 FPS: 1.004 PSNR: 27.33 QRCRv1.0: 0 SSIM: 0.808 Subjective score: 5.353 |