Real-World Knowledge-Guided Change Data Synthesis for Remote Sensing
KnowChange uses pretrained vision-language models to decide where and how remote-sensing scenes should plausibly change.
The paper says this replaces handcrafted change rules that limit class transitions and flexibility. Its framework combines knowledge-guided change simulation with generalizable synthesis models to produce varied change types. In experiments, KnowChange-generated data beat existing synthetic datasets for synthetic-to-real transfer and data augmentation, even at smaller scale. HF Daily Papers' note
The paper says this replaces handcrafted change rules that limit class transitions and flexibility. Its framework combines knowledge-guided change simulation with generalizable synthesis models to produce varied change types. In experiments, KnowChange-generated data beat existing synthetic datasets for synthetic-to-real transfer and data augmentation, even at smaller scale. HF Daily Papers' note
score 4