{"id":18842387,"url":"https://github.com/lusinlu/gradient-variance-loss","last_synced_at":"2025-06-20T08:08:40.408Z","repository":{"id":39335602,"uuid":"399741512","full_name":"lusinlu/gradient-variance-loss","owner":"lusinlu","description":"Code of the ICASSP 2022 paper \"Gradient Variance Loss for Structure Enhanced Super-Resolution\"","archived":false,"fork":false,"pushed_at":"2022-02-27T16:05:14.000Z","size":209,"stargazers_count":31,"open_issues_count":1,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-04-14T07:47:08.874Z","etag":null,"topics":["deep-learning","loss","loss-functions","machine-learning","model-optimization","super-resolution","training","upscaling"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/lusinlu.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null}},"created_at":"2021-08-25T08:14:07.000Z","updated_at":"2025-03-23T03:50:59.000Z","dependencies_parsed_at":"2022-09-20T03:20:42.168Z","dependency_job_id":null,"html_url":"https://github.com/lusinlu/gradient-variance-loss","commit_stats":null,"previous_names":["lusinlu/gradient_variance_loss"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/lusinlu/gradient-variance-loss","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lusinlu%2Fgradient-variance-loss","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lusinlu%2Fgradient-variance-loss/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lusinlu%2Fgradient-variance-loss/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lusinlu%2Fgradient-variance-loss/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/lusinlu","download_url":"https://codeload.github.com/lusinlu/gradient-variance-loss/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lusinlu%2Fgradient-variance-loss/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":260907160,"owners_count":23080613,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["deep-learning","loss","loss-functions","machine-learning","model-optimization","super-resolution","training","upscaling"],"created_at":"2024-11-08T02:54:33.393Z","updated_at":"2025-06-20T08:08:35.394Z","avatar_url":"https://github.com/lusinlu.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Gradient Variance Loss\n[![report](https://img.shields.io/badge/arxiv-report-red)](https://arxiv.org/abs/2202.00997)\n\n[ICASSP 2022] Official implementation of the Gradient Variance loss presented in the paper [paper](https://arxiv.org/abs/2202.00997) \"Gradient Variance Loss for Structure-Enhanced Image Super-Resolution\".\n\n## Requirements\nfor installing required packages run\n` pip install -r requirements.txt`\n\n## Usage\nTo train the VDSR model with the gradient variance loss run the following command\n\n` python train.py --dataroot [path to DIV2K dataset] --cuda`\n\n\n## Introduction\n\"Gradient Variance Loss for Structure-Enhanced Image Super-Resolution\"\n\nBy Lusine Abrahamyan, Anh Minh Truong, Wilfried Philips and Nikos Deligiannis.\n### Approach\nWe observe that gradient maps of images generated\nby the models trained with the L1/L2 losses have significantly lower variance than the gradient maps of the original\nhigh-resolution images. \n\nIn this work, we introduce a structure-enhancing loss\nfunction, coined Gradient Variance (GV) loss, to minimize the difference between the variances of predicted and original gradient maps and generate\ntextures with perceptual-pleasant details.\n\n### Performance\nPublic benchmark test results and DIV2K validation results (PSNR(dB) / SSIM).\n\n\u003cimg src=\"https://github.com/lusinlu/gradient_variance_loss/blob/main/table_results.png\" width=\"750\" height=\"250\"\u003e\n\n## Citation\nIf you find the code useful for your research, please consider citing our works\n\n```\n@article{abrahamyangvloss,\n  title={Gradient Variance Loss for Structure-Enhanced Image Super-Resolution},\n  author={Lusine, Abrahamyan and  Anh Minh, Truong and  Wilfried, Philips and Nikos, Deligiannis},\n  journal={Proceedings of the International Conference on Acoustics, Speech, and Signal Processing (ICASSP)},\n  publisher = {IEEE},\n  year={2022}\n}\n```\n\n## Acknowledgement\nCodes for the VDSR model are from [pytorch-vdsr](https://github.com/twtygqyy/pytorch-vdsr).\n\n\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flusinlu%2Fgradient-variance-loss","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Flusinlu%2Fgradient-variance-loss","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flusinlu%2Fgradient-variance-loss/lists"}