AIM 2019 Challenge on Real-World Image Super-Resolution: Methods and Results

Lugmayr, Andreas; Joon, Nam Hyung; Won, Yu Seung; Kim, Guisik; Kwon, Dokyeong; Hsu, Chih–Chung; Lin, Chia-Hsiang; Huang, Yuanfei; Sun, Xingming; Lu, Wen; Li, Jie; Danelljan, Martin; Gao, Xinbo; Bell-Kligler, Sefi; Shocher, Assaf; Irani, Michal; Timofte, Radu; Fritsche, Manuel; Gu, Shuhang; Purohit, Kuldeep; Kandula, Praveen; Suin, Maitreya; Rajagoapalan, A N

doi:10.1109/iccvw.2019.00442

Cited by 115 publications

(99 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, we also apply our idea of frequency separation to the SR model, which allows it to match the target distribution more closely. Our multiple experiments with artificial and natural corruptions demonstrate the effectiveness of our approach for real-world SR. We not only beat the state-of-the-art methods in these experiments, but also won the AIM 2019 Challenge on Real World Super-Resolution [27].…”

Section: Resultsmentioning

confidence: 98%

“…Furthermore, we also participated in the AIM 2019 Challenge on Real World Super-Resolution [27]. In this challenge 2650 corrupted source images and 800 clean target images are provided for training.…”

Section: Esrgan and Esrgan-fsmentioning

confidence: 99%

“…The AIM 2019 Challenge We also participated in the AIM 2019 Real World Super-Resolution Challenge [27]. Thereby, the SR (×4) images are supposed to either match the source domain in Same Domain Real World Super-Resolution (SDSR) or a clean target domain in Target Domain Real World Super-Resolution (TDSR).…”

Section: Experiments On Real-world Imagesmentioning

confidence: 99%

“…For both SDSR and TDSR our method won the challenge by achieving the lowest MOS. More information on the evaluation and competing methods can be found in the challenge report [27].…”

Section: Experiments On Real-world Imagesmentioning

confidence: 99%

“…We only provide qualitative evaluation in this case since no GT is available. Finally, we also participated in the AIM 2019 Challenge on Real World Super-Resolution [27] associated with the AIM workshop at ICCV 2019. Our method won the first place in both tracks for source domain and target domain.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Frequency Separation for Real-World Super-Resolution

Fritsche

Timofte

2019

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)

Self Cite

229

187

View full text Add to dashboard Cite

Most of the recent literature on image super-resolution (SR) assumes the availability of training data in the form of paired low resolution (LR) and high resolution (HR) images or the knowledge of the downgrading operator (usually bicubic downscaling). While the proposed methods perform well on standard benchmarks, they often fail to produce convincing results in real-world settings. This is because real-world images can be subject to corruptions such as sensor noise, which are severely altered by bicubic downscaling. Therefore, the models never see a real-world image during training, which limits their generalization capabilities. Moreover, it is cumbersome to collect paired LR and HR images in the same source domain.To address this problem, we propose DSGAN to introduce natural image characteristics in bicubically downscaled images. It can be trained in an unsupervised fashion on HR images, thereby generating LR images with the same characteristics as the original images. We then use the generated data to train a SR model, which greatly improves its performance on real-world images. Furthermore, we propose to separate the low and high image frequencies and treat them differently during training. Since the low frequencies are preserved by downsampling operations, we only require adversarial training to modify the high frequencies. This idea is applied to our DSGAN model as well as the SR model. We demonstrate the effectiveness of our method in several experiments through quantitative and qualitative analysis. Our solution is the winner of the AIM Challenge on Real World SR at ICCV 2019. arXiv:1911.07850v1 [eess.IV] 18 Nov 2019 1 Our code and models are publicly available at https://github. com/ManuelFritsche/real-world-sr

show abstract

Section: Resultsmentioning

confidence: 98%

Section: Esrgan and Esrgan-fsmentioning

confidence: 99%

Section: Experiments On Real-world Imagesmentioning

confidence: 99%

“…For both SDSR and TDSR our method won the challenge by achieving the lowest MOS. More information on the evaluation and competing methods can be found in the challenge report [27].…”

Section: Experiments On Real-world Imagesmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Frequency Separation for Real-World Super-Resolution

Fritsche

Timofte

2019

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)

Self Cite

229

187

View full text Add to dashboard Cite

show abstract

SESAME: Semantic Editing of Scenes by Adding, Manipulating or Erasing Objects

Ntavelis

Romero

Kastanis

et al. 2020

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

SRFlow: Learning the Super-Resolution Space with Normalizing Flow

Lugmayr

Danelljan

Gool

et al. 2020

Lecture Notes in Computer Science

Self Cite

276

250

View full text Add to dashboard Cite

Super-resolution is an ill-posed problem, since it allows for multiple predictions for a given low-resolution image. This fundamental fact is largely ignored by state-of-the-art deep learning based approaches. These methods instead train a deterministic mapping using combinations of reconstruction and adversarial losses. In this work, we therefore propose SRFlow: a normalizing flow based super-resolution method capable of learning the conditional distribution of the output given the lowresolution input. Our model is trained in a principled manner using a single loss, namely the negative log-likelihood. SRFlow therefore directly accounts for the ill-posed nature of the problem, and learns to predict diverse photo-realistic high-resolution images. Moreover, we utilize the strong image posterior learned by SRFlow to design flexible image manipulation techniques, capable of enhancing super-resolved images by, e.g., transferring content from other images. We perform extensive experiments on faces, as well as on super-resolution in general. SRFlow outperforms state-of-the-art GAN-based approaches in terms of both PSNR and perceptual quality metrics, while allowing for diversity through the exploration of the space of super-resolved solutions. Code: git.io/Jfpyu.

show abstract

AIM 2019 Challenge on Real-World Image Super-Resolution: Methods and Results

Cited by 115 publications

References 34 publications

Frequency Separation for Real-World Super-Resolution

Frequency Separation for Real-World Super-Resolution

SESAME: Semantic Editing of Scenes by Adding, Manipulating or Erasing Objects

SRFlow: Learning the Super-Resolution Space with Normalizing Flow

Contact Info

Product

Resources

About