Reproduce “Referring Change Detection in Remote Sensing Imagery” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yilmaz Korkmaz, Jay N. Paranjape, Celso M. de Melo, Vishal M. Patel
Researchers from Johns Hopkins University and DEVCOM U.S. Army Research Laboratory developed a framework for Referring Change Detection (RCD) in remote sensing imagery, enabling users to identify specific changes via natural language prompts. Their RCDNet model, combined with a novel diffusion-based synthetic data generation pipeline (RCDGen), achieves state-of-the-art performance across semantic, binary, and cross-domain change detection tasks.
Key insights
- Integrating natural language guidance enables highly flexible and user-specific change detection, overcoming the rigid class definitions of traditional SCD.
- A diffusion-based synthetic data generation pipeline can effectively address data scarcity and imbalance, substantially improving model generalization, especially in cross-domain scenarios.
- A lightweight Mamba-based architecture, when combined with cross-attention to text embeddings, efficiently processes visual and linguistic information for precise referring change localization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.