Papers › Flexible Style Image Super-Resolution using Conditional Objective

Flexible Style Image Super-Resolution using Conditional Objective

13 Jan 2022arXiv:2201.04898archive 2025-07-28

Seung Ho Park, Young Su Moon, Nam Ik Cho

Recent studies have significantly enhanced the performance of single-image super-resolution (SR) using convolutional neural networks (CNNs). While there can be many high-resolution (HR) solutions for a given input, most existing CNN-based methods do not explore alternative solutions during the inference. A typical approach to obtaining alternative SR results is to train multiple SR models with different loss weightings and exploit the combination of these models. Instead of using multiple models, we present a more efficient method to train a single adjustable SR model on various combinations of losses by taking advantage of multi-task learning. Specifically, we optimize an SR model with a conditional objective during training, where the objective is a weighted sum of multiple perceptual losses at different feature levels. The weights vary according to given conditions, and the set of weights is defined as a style controller. Also, we present an architecture appropriate for this training scheme, which is the Residual-in-Residual Dense Block equipped with spatial feature transformation layers. At the inference phase, our trained model can generate locally different outputs conditioned on the style control map. Extensive experiments show that the proposed SR model produces various desirable reconstructions without artifacts and yields comparable quantitative performance to state-of-the-art SR methods.

PaperPDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

seungho-snu/fxsr officialmentioned in paperpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Image Super-ResolutionMulti-Task LearningSemantic SegmentationSuper-Resolution

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Super-Resolution BSD100 - 4x upscaling FxSR-PD t=0.0 LPIPS 0.3433 #58 of 71 Archive leaderboard report
Image Super-Resolution BSD100 - 4x upscaling FxSR-PD t=0.0 PSNR 26.38 #58 of 71 Archive leaderboard report
Image Super-Resolution BSD100 - 4x upscaling FxSR-PD t=0.0 SSIM 0.738 #58 of 71 Archive leaderboard report
Image Super-Resolution BSD100 - 4x upscaling FxSR-PD t=0.8 LPIPS 0.1572 #67 of 71 Archive leaderboard report
Image Super-Resolution BSD100 - 4x upscaling FxSR-PD t=0.8 PSNR 24.77 #67 of 71 Archive leaderboard report
Image Super-Resolution BSD100 - 4x upscaling FxSR-PD t=0.8 SSIM 0.6817 #67 of 71 Archive leaderboard report
Image Super-Resolution BSD100 - 8x upscaling FxSR-PD t=0.0 DISTS 0.2753 #5 of 6 Archive leaderboard report
Image Super-Resolution BSD100 - 8x upscaling FxSR-PD t=0.0 LPIPS 0.5079 #5 of 6 Archive leaderboard report
Image Super-Resolution BSD100 - 8x upscaling FxSR-PD t=0.0 LRPSNR 47.12 #5 of 6 Archive leaderboard report
Image Super-Resolution BSD100 - 8x upscaling FxSR-PD t=0.0 NIQE 5.49 #5 of 6 Archive leaderboard report
Image Super-Resolution BSD100 - 8x upscaling FxSR-PD t=0.0 PSNR 23.6 #5 of 6 Archive leaderboard report
Image Super-Resolution BSD100 - 8x upscaling FxSR-PD t=0.0 SSIM 0.5728 #5 of 6 Archive leaderboard report
Image Super-Resolution BSD100 - 8x upscaling FxSR-PD t=0.8 DISTS 0.1972 #6 of 6 Archive leaderboard report
Image Super-Resolution BSD100 - 8x upscaling FxSR-PD t=0.8 LPIPS 0.3129 #6 of 6 Archive leaderboard report
Image Super-Resolution BSD100 - 8x upscaling FxSR-PD t=0.8 LRPSNR 42.41 #6 of 6 Archive leaderboard report
Image Super-Resolution BSD100 - 8x upscaling FxSR-PD t=0.8 NIQE 4.58 #6 of 6 Archive leaderboard report
Image Super-Resolution BSD100 - 8x upscaling FxSR-PD t=0.8 PSNR 21.93 #6 of 6 Archive leaderboard report
Image Super-Resolution BSD100 - 8x upscaling FxSR-PD t=0.8 SSIM 0.5039 #6 of 6 Archive leaderboard report
Image Super-Resolution DIV2K val - 4x upscaling FxSR-PD t=0.8 DISTS 0.0513 #3 of 21 Archive leaderboard report
Image Super-Resolution DIV2K val - 4x upscaling FxSR-PD t=0.8 LPIPS 0.1028 #3 of 21 Archive leaderboard report
Image Super-Resolution DIV2K val - 4x upscaling FxSR-PD t=0.8 LRPSNR 50.54 #3 of 21 Archive leaderboard report
Image Super-Resolution DIV2K val - 4x upscaling FxSR-PD t=0.8 NIQE 2.81 #3 of 21 Archive leaderboard report
Image Super-Resolution DIV2K val - 4x upscaling FxSR-PD t=0.8 PSNR 27.51 #3 of 21 Archive leaderboard report
Image Super-Resolution DIV2K val - 4x upscaling FxSR-PD t=0.8 SSIM 0.789 #3 of 21 Archive leaderboard report
Image Super-Resolution DIV2K val - 4x upscaling FxSR-PD t=0.0 DISTS 0.1169 #10 of 21 Archive leaderboard report
Image Super-Resolution DIV2K val - 4x upscaling FxSR-PD t=0.0 LPIPS 0.239 #10 of 21 Archive leaderboard report
Image Super-Resolution DIV2K val - 4x upscaling FxSR-PD t=0.0 LRPSNR 53.3 #10 of 21 Archive leaderboard report
Image Super-Resolution DIV2K val - 4x upscaling FxSR-PD t=0.0 NIQE 4.11 #10 of 21 Archive leaderboard report
Image Super-Resolution DIV2K val - 4x upscaling FxSR-PD t=0.0 PSNR 29.24 #10 of 21 Archive leaderboard report
Image Super-Resolution DIV2K val - 4x upscaling FxSR-PD t=0.0 SSIM 0.8383 #10 of 21 Archive leaderboard report
Image Super-Resolution DIV2K val - 8x upscaling FxSR-PD t=0.8 DISTS 0.119 #1 of 2 Archive leaderboard report
Image Super-Resolution DIV2K val - 8x upscaling FxSR-PD t=0.8 LPIPS 0.2403 #1 of 2 Archive leaderboard report
Image Super-Resolution DIV2K val - 8x upscaling FxSR-PD t=0.8 LRPSNR 42.66 #1 of 2 Archive leaderboard report
Image Super-Resolution DIV2K val - 8x upscaling FxSR-PD t=0.8 NIQE 3.61 #1 of 2 Archive leaderboard report
Image Super-Resolution DIV2K val - 8x upscaling FxSR-PD t=0.8 PSNR 23.56 #1 of 2 Archive leaderboard report
Image Super-Resolution DIV2K val - 8x upscaling FxSR-PD t=0.8 SSIM 0.6241 #1 of 2 Archive leaderboard report
Image Super-Resolution DIV2K val - 8x upscaling FxSR-PD t=0.0 DISTS 0.1953 #2 of 2 Archive leaderboard report
Image Super-Resolution DIV2K val - 8x upscaling FxSR-PD t=0.0 LPIPS 0.3857 #2 of 2 Archive leaderboard report
Image Super-Resolution DIV2K val - 8x upscaling FxSR-PD t=0.0 LRPSNR 46.96 #2 of 2 Archive leaderboard report
Image Super-Resolution DIV2K val - 8x upscaling FxSR-PD t=0.0 NIQE 4.41 #2 of 2 Archive leaderboard report
Image Super-Resolution DIV2K val - 8x upscaling FxSR-PD t=0.0 PSNR 25.6 #2 of 2 Archive leaderboard report
Image Super-Resolution DIV2K val - 8x upscaling FxSR-PD t=0.0 SSIM 0.6989 #2 of 2 Archive leaderboard report
Image Super-Resolution General100 - 4x upscaling FxSR-PD t=0.8 DISTS 0.0831 #2 of 3 Archive leaderboard report
Image Super-Resolution General100 - 4x upscaling FxSR-PD t=0.8 LPIPS 0.0784 #2 of 3 Archive leaderboard report
Image Super-Resolution General100 - 4x upscaling FxSR-PD t=0.8 LRPSNR 49.82 #2 of 3 Archive leaderboard report
Image Super-Resolution General100 - 4x upscaling FxSR-PD t=0.8 NIQE 4.54 #2 of 3 Archive leaderboard report
Image Super-Resolution General100 - 4x upscaling FxSR-PD t=0.8 PSNR 28.44 #2 of 3 Archive leaderboard report
Image Super-Resolution General100 - 4x upscaling FxSR-PD t=0.8 SSIM 0.8229 #2 of 3 Archive leaderboard report
Image Super-Resolution General100 - 4x upscaling FxSR-PD t=0.0 DISTS 0.1205 #3 of 3 Archive leaderboard report
Image Super-Resolution General100 - 4x upscaling FxSR-PD t=0.0 LPIPS 0.1519 #3 of 3 Archive leaderboard report
Image Super-Resolution General100 - 4x upscaling FxSR-PD t=0.0 LRPSNR 52.22 #3 of 3 Archive leaderboard report
Image Super-Resolution General100 - 4x upscaling FxSR-PD t=0.0 NIQE 6.05 #3 of 3 Archive leaderboard report
Image Super-Resolution General100 - 4x upscaling FxSR-PD t=0.0 PSNR 29.94 #3 of 3 Archive leaderboard report
Image Super-Resolution General100 - 4x upscaling FxSR-PD t=0.0 SSIM 0.8629 #3 of 3 Archive leaderboard report
Image Super-Resolution General100 - 8x upscaling FxSR-PD t=0.8 DISTS 0.1716 #1 of 2 Archive leaderboard report
Image Super-Resolution General100 - 8x upscaling FxSR-PD t=0.8 LPIPS 0.2058 #1 of 2 Archive leaderboard report
Image Super-Resolution General100 - 8x upscaling FxSR-PD t=0.8 LRPSNR 41.36 #1 of 2 Archive leaderboard report
Image Super-Resolution General100 - 8x upscaling FxSR-PD t=0.8 NIQE 5.46 #1 of 2 Archive leaderboard report
Image Super-Resolution General100 - 8x upscaling FxSR-PD t=0.8 PSNR 24 #1 of 2 Archive leaderboard report
Image Super-Resolution General100 - 8x upscaling FxSR-PD t=0.8 SSIM 0.6534 #1 of 2 Archive leaderboard report
Image Super-Resolution General100 - 8x upscaling FxSR-PD t=0.0 DISTS 0.2134 #2 of 2 Archive leaderboard report
Image Super-Resolution General100 - 8x upscaling FxSR-PD t=0.0 LPIPS 0.2924 #2 of 2 Archive leaderboard report
Image Super-Resolution General100 - 8x upscaling FxSR-PD t=0.0 LRPSNR 44.28 #2 of 2 Archive leaderboard report
Image Super-Resolution General100 - 8x upscaling FxSR-PD t=0.0 NIQE 6.09 #2 of 2 Archive leaderboard report
Image Super-Resolution General100 - 8x upscaling FxSR-PD t=0.0 PSNR 25.42 #2 of 2 Archive leaderboard report
Image Super-Resolution General100 - 8x upscaling FxSR-PD t=0.0 SSIM 0.7097 #2 of 2 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Batch NormalizationConcatenated Skip ConnectionConvolutionDense BlockReLU

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections