{"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/curriculum-model-adaptation-with-synthetic","title":"Curriculum Model Adaptation with Synthetic and Real Data for Semantic Foggy Scene Understanding","arxiv_id":"1901.01415","date":"2019-01-05","proceeding":null,"authors":["Dengxin Dai","Christos Sakaridis","Simon Hecker","Luc van Gool"],"abstract":"This work addresses the problem of semantic scene understanding under fog.\nAlthough marked progress has been made in semantic scene understanding, it is\nmainly concentrated on clear-weather scenes. Extending semantic segmentation\nmethods to adverse weather conditions such as fog is crucial for outdoor\napplications. In this paper, we propose a novel method, named Curriculum Model\nAdaptation (CMAda), which gradually adapts a semantic segmentation model from\nlight synthetic fog to dense real fog in multiple steps, using both labeled\nsynthetic foggy data and unlabeled real foggy data. The method is based on the\nfact that the results of semantic segmentation in moderately adverse conditions\n(light fog) can be bootstrapped to solve the same problem in highly adverse\nconditions (dense fog). CMAda is extensible to other adverse conditions and\nprovides a new paradigm for learning with synthetic data and unlabeled real\ndata. In addition, we present three other main stand-alone contributions: 1) a\nnovel method to add synthetic fog to real, clear-weather scenes using semantic\ninput; 2) a new fog density estimator; 3) a novel fog densification method to\ndensify the fog in real foggy scenes without using depth; and 4) the Foggy\nZurich dataset comprising 3808 real foggy images, with pixel-level semantic\nannotations for 40 images under dense fog. Our experiments show that 1) our fog\nsimulation and fog density estimator outperform their state-of-the-art\ncounterparts with respect to the task of semantic foggy scene understanding\n(SFSU); 2) CMAda improves the performance of state-of-the-art models for SFSU\nsignificantly, benefiting both from our synthetic and real foggy data. The\ndatasets and code are available at the project website.","url_abs":"http://arxiv.org/abs/1901.01415v2","url_pdf":"http://arxiv.org/pdf/1901.01415v2.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":"curriculum-model-adaptation-with-synthetic","repo_url":"https://github.com/sohyun-l/fifo","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"domain-adaptation","task_name":"Domain Adaptation"},{"task_slug":"scene-understanding","task_name":"Scene Understanding"},{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/domain-adaptation-on-cityscapes-to-1","task":"Domain Adaptation","dataset":"Cityscapes-to-FoggyDriving","model":"CMAda3+","rank_in_archive_order":5,"of":5,"metrics":{"mIoU":"49.8"},"uses_additional_data":false},{"leaderboard":"/sota/domain-adaptation-on-cityscapes-to","task":"Domain Adaptation","dataset":"Cityscapes-to-FoggyZurich","model":"CMAda3+","rank_in_archive_order":6,"of":6,"metrics":{"mIoU":"46.8"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1901.01415","atlas_url":"https://app.syntology.ai/?focus=1901.01415","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}