Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
The paper introduces OmniCD, a foundational framework for remote sensing image change detection that leverages multimodal semantic guidance from image and text prompts. It unifies binary CD to zero-shot semantic change understanding through a hierarchical scene retrieval module and a change detection module, enhanced by style disentanglement. Experiments on the newly introduced RSITCD dataset (300K+ image-text pairs) demonstrate state-of-the-art performance and strong cross-domain adaptability.
Remote sensing change detection gets a unified framework that understands both images and text, unlocking zero-shot capabilities and outperforming existing methods.
Change detection (CD) in remote sensing is vital for applications such as urban monitoring and disaster assessment, yet traditional methods struggle with generalization across diverse scenarios. We present OmniCD, a foundational framework that unifies and enhances remote sensing CD through multimodal semantic guidance. OmniCD incorporates image and text prompts -- such as textual descriptions, semantic maps, and geospatial metadata -- into a unified architecture, supporting tasks from binary CD to zero-shot semantic change understanding. The framework integrates a hierarchical scene retrieval module and a change detection module, reinforced by a style disentanglement mechanism for improved cross-domain robustness. We further introduce RSITCD, a large-scale multimodal dataset with 300K+ annotated image-text pairs. Extensive experiments show that OmniCD achieves state-of-the-art performance across benchmarks, demonstrating strong adaptability and setting a solid foundation for general-purpose CD systems in remote sensing.