Abstract
Change detection (CD) is a fundamental task for monitoring and analyzing landcover dynamics. While recent high performance models and high quality datasetshave significantly advanced the field, a critical limitation persists. Currentmodels typically acquire limited knowledge from single-type annotated data andcannot concurrently leverage diverse binary change detection (BCD) and semanticchange detection (SCD) datasets. This constraint leads to poor generalizationand limited versatility. The recent advancements in Multimodal Large LanguageModels (MLLMs) introduce new possibilities for a unified CD framework. Weleverage the language priors and unification capabilities of MLLMs to developUniChange, the first MLLM-based unified change detection model. UniChangeintegrates generative language abilities with specialized CD functionalities.Our model successfully unifies both BCD and SCD tasks through the introductionof three special tokens: [T1], [T2], and [CHANGE]. Furthermore, UniChangeutilizes text prompts to guide the identification of change categories,eliminating the reliance on predefined classification heads. This design allowsUniChange to effectively acquire knowledge from multi-source datasets, evenwhen their class definitions conflict. Experiments on four public benchmarks(WHU-CD, S2Looking, LEVIR-CD+, and SECOND) demonstrate SOTA performance,achieving IoU scores of 90.41, 53.04, 78.87, and 57.62, respectively,surpassing all previous methods. The code is available athttps://github.com/Erxucomeon/UniChange.