{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/flexible-style-image-super-resolution-using","title":"Flexible Style Image Super-Resolution using Conditional Objective","arxiv_id":"2201.04898","date":"2022-01-13","proceeding":null,"authors":["Seung Ho Park","Young Su Moon","Nam Ik Cho"],"abstract":"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.","url_abs":"https://arxiv.org/abs/2201.04898v3","url_pdf":"https://arxiv.org/pdf/2201.04898v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"flexible-style-image-super-resolution-using","repo_url":"https://github.com/seungho-snu/fxsr","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"image-super-resolution","task_name":"Image Super-Resolution"},{"task_slug":"multi-task-learning","task_name":"Multi-Task Learning"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"super-resolution","task_name":"Super-Resolution"}],"methods":[{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"concatenated-skip-connection","method_name":"Concatenated Skip Connection"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dense-block","method_name":"Dense Block"},{"method_slug":"relu","method_name":"ReLU"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-super-resolution-on-bsd100-4x-upscaling","task":"Image Super-Resolution","dataset":"BSD100 - 4x upscaling","model":"FxSR-PD t=0.0","rank_in_archive_order":58,"of":71,"metrics":{"LPIPS":"0.3433","PSNR":"26.38","SSIM":"0.738"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-bsd100-4x-upscaling","task":"Image Super-Resolution","dataset":"BSD100 - 4x upscaling","model":"FxSR-PD t=0.8","rank_in_archive_order":67,"of":71,"metrics":{"LPIPS":"0.1572","PSNR":"24.77","SSIM":"0.6817"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-bsd100-8x-upscaling","task":"Image Super-Resolution","dataset":"BSD100 - 8x upscaling","model":"FxSR-PD t=0.0","rank_in_archive_order":5,"of":6,"metrics":{"DISTS":"0.2753","LPIPS":"0.5079","LRPSNR":"47.12","NIQE":"5.49","PSNR":"23.6","SSIM":"0.5728"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-bsd100-8x-upscaling","task":"Image Super-Resolution","dataset":"BSD100 - 8x upscaling","model":"FxSR-PD t=0.8","rank_in_archive_order":6,"of":6,"metrics":{"DISTS":"0.1972","LPIPS":"0.3129","LRPSNR":"42.41","NIQE":"4.58","PSNR":"21.93","SSIM":"0.5039"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-div2k-val-4x","task":"Image Super-Resolution","dataset":"DIV2K val - 4x upscaling","model":"FxSR-PD t=0.8","rank_in_archive_order":3,"of":21,"metrics":{"DISTS":"0.0513","LPIPS":"0.1028","LRPSNR":"50.54","NIQE":"2.81","PSNR":"27.51","SSIM":"0.789"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-div2k-val-4x","task":"Image Super-Resolution","dataset":"DIV2K val - 4x upscaling","model":"FxSR-PD t=0.0","rank_in_archive_order":10,"of":21,"metrics":{"DISTS":"0.1169","LPIPS":"0.239","LRPSNR":"53.3","NIQE":"4.11","PSNR":"29.24","SSIM":"0.8383"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-div2k-val-8x","task":"Image Super-Resolution","dataset":"DIV2K val - 8x upscaling","model":"FxSR-PD t=0.8","rank_in_archive_order":1,"of":2,"metrics":{"DISTS":"0.119","LPIPS":"0.2403","LRPSNR":"42.66","NIQE":"3.61","PSNR":"23.56","SSIM":"0.6241"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-div2k-val-8x","task":"Image Super-Resolution","dataset":"DIV2K val - 8x upscaling","model":"FxSR-PD t=0.0","rank_in_archive_order":2,"of":2,"metrics":{"DISTS":"0.1953","LPIPS":"0.3857","LRPSNR":"46.96","NIQE":"4.41","PSNR":"25.6","SSIM":"0.6989"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-general100-4x","task":"Image Super-Resolution","dataset":"General100 - 4x upscaling","model":"FxSR-PD t=0.8","rank_in_archive_order":2,"of":3,"metrics":{"DISTS":"0.0831","LPIPS":"0.0784","LRPSNR":"49.82","NIQE":"4.54","PSNR":"28.44","SSIM":"0.8229"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-general100-4x","task":"Image Super-Resolution","dataset":"General100 - 4x upscaling","model":"FxSR-PD t=0.0","rank_in_archive_order":3,"of":3,"metrics":{"DISTS":"0.1205","LPIPS":"0.1519","LRPSNR":"52.22","NIQE":"6.05","PSNR":"29.94","SSIM":"0.8629"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-general100-8x","task":"Image Super-Resolution","dataset":"General100 - 8x upscaling","model":"FxSR-PD t=0.8","rank_in_archive_order":1,"of":2,"metrics":{"DISTS":"0.1716","LPIPS":"0.2058","LRPSNR":"41.36","NIQE":"5.46","PSNR":"24","SSIM":"0.6534"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-general100-8x","task":"Image Super-Resolution","dataset":"General100 - 8x upscaling","model":"FxSR-PD t=0.0","rank_in_archive_order":2,"of":2,"metrics":{"DISTS":"0.2134","LPIPS":"0.2924","LRPSNR":"44.28","NIQE":"6.09","PSNR":"25.42","SSIM":"0.7097"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2201.04898","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}