{"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/schanger-change-detection-from-a-semantic","title":"SChanger: Change Detection from a Semantic Change and Spatial Consistency Perspective","arxiv_id":"2503.20734","date":"2025-03-26","proceeding":null,"authors":["Ziyu Zhou","Keyan Hu","Yutian Fang","Xiaoping Rui"],"abstract":"Change detection is a key task in Earth observation applications. Recently, deep learning methods have demonstrated strong performance and widespread application. However, change detection faces data scarcity due to the labor-intensive process of accurately aligning remote sensing images of the same area, which limits the performance of deep learning algorithms. To address the data scarcity issue, we develop a fine-tuning strategy called the Semantic Change Network (SCN). We initially pre-train the model on single-temporal supervised tasks to acquire prior knowledge of instance feature extraction. The model then employs a shared-weight Siamese architecture and extended Temporal Fusion Module (TFM) to preserve this prior knowledge and is fine-tuned on change detection tasks. The learned semantics for identifying all instances is changed to focus on identifying only the changes. Meanwhile, we observe that the locations of changes between the two images are spatially identical, a concept we refer to as spatial consistency. We introduce this inductive bias through an attention map that is generated by large-kernel convolutions and applied to the features from both time points. This enhances the modeling of multi-scale changes and helps capture underlying relationships in change detection semantics. We develop a binary change detection model utilizing these two strategies. The model is validated against state-of-the-art methods on six datasets, surpassing all benchmark methods and achieving F1 scores of 92.87%, 86.43%, 68.95%, 97.62%, 84.58%, and 93.20% on the LEVIR-CD, LEVIR-CD+, S2Looking, CDD, SYSU-CD, and WHU-CD datasets, respectively.","url_abs":"https://arxiv.org/abs/2503.20734v1","url_pdf":"https://arxiv.org/pdf/2503.20734v1.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":"schanger-change-detection-from-a-semantic","repo_url":"https://github.com/zhouziyu-cn/SChanger","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"change-detection","task_name":"Change Detection"}],"methods":[{"method_slug":"attention","method_name":"Attention"},{"method_slug":"focus","method_name":"Focus"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/change-detection-on-cdd-dataset-season-1","task":"Change Detection","dataset":"CDD Dataset (season-varying)","model":"SChanger-base","rank_in_archive_order":7,"of":18,"metrics":{"F1-Score":"97.62"},"uses_additional_data":false},{"leaderboard":"/sota/change-detection-on-cdd-dataset-season-1","task":"Change Detection","dataset":"CDD Dataset (season-varying)","model":"SChanger-small","rank_in_archive_order":11,"of":18,"metrics":{"F1-Score":"95.75"},"uses_additional_data":false},{"leaderboard":"/sota/change-detection-on-levir","task":"Change Detection","dataset":"LEVIR+","model":"SChanger-base","rank_in_archive_order":2,"of":9,"metrics":{"F1":"86.43"},"uses_additional_data":false},{"leaderboard":"/sota/change-detection-on-levir","task":"Change Detection","dataset":"LEVIR+","model":"SChanger-small","rank_in_archive_order":3,"of":9,"metrics":{"F1":"86.20"},"uses_additional_data":false},{"leaderboard":"/sota/change-detection-on-levir-cd","task":"Change Detection","dataset":"LEVIR-CD","model":"SChanger-base","rank_in_archive_order":1,"of":28,"metrics":{"F1":"92.87"},"uses_additional_data":false},{"leaderboard":"/sota/change-detection-on-levir-cd","task":"Change Detection","dataset":"LEVIR-CD","model":"SChanger-small","rank_in_archive_order":4,"of":28,"metrics":{"F1":"92.45"},"uses_additional_data":false},{"leaderboard":"/sota/change-detection-on-s2looking","task":"Change Detection","dataset":"S2Looking","model":"SChanger-base","rank_in_archive_order":1,"of":11,"metrics":{"F1-Score":"68.95"},"uses_additional_data":false},{"leaderboard":"/sota/change-detection-on-s2looking","task":"Change Detection","dataset":"S2Looking","model":"SChanger-small","rank_in_archive_order":2,"of":11,"metrics":{"F1-Score":"68.20"},"uses_additional_data":false},{"leaderboard":"/sota/change-detection-on-sysu-cd","task":"Change Detection","dataset":"SYSU-CD","model":"SChanger-small","rank_in_archive_order":1,"of":12,"metrics":{"F1":"84.58"},"uses_additional_data":false},{"leaderboard":"/sota/change-detection-on-sysu-cd","task":"Change Detection","dataset":"SYSU-CD","model":"SChanger-base","rank_in_archive_order":2,"of":12,"metrics":{"F1":"84.17"},"uses_additional_data":false},{"leaderboard":"/sota/change-detection-on-whu-cd","task":"Change Detection","dataset":"WHU-CD","model":"SChanger-base","rank_in_archive_order":6,"of":22,"metrics":{"F1":"93.20"},"uses_additional_data":false},{"leaderboard":"/sota/change-detection-on-whu-cd","task":"Change Detection","dataset":"WHU-CD","model":"SChanger-small","rank_in_archive_order":7,"of":22,"metrics":{"F1":"93.15"},"uses_additional_data":false}],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}