Real-World Knowledge-Guided Change Data Synthesis for Remote Sensing
Researchers introduce KnowChange, a framework that uses pretrained vision-language models to synthesize realistic change-detection training data for remote sensing.
KnowChange is a knowledge-guided change data synthesis framework that leverages pretrained vision-language models to reason about plausible change locations and class transitions from pre-change scenes and desired change types. It addresses the limited class-transition coverage and inflexibility of handcrafted rule-based synthesis methods, enabling diverse change types in a unified pipeline. Experiments show KnowChange-generated data outperforms existing synthetic datasets in both synthetic-to-real transfer and synthetic data augmentation, despite compact generation scale.
- Uses pretrained vision-language models as knowledge sources for change simulation
- Overcomes limited class-transition coverage of rule-based synthesis
- Outperforms prior synthetic datasets in transfer and augmentation tasks
- Integrates into existing synthesis pipelines to boost downstream utility
- Tags
- change-detectiondata-synthesispretrained-modelsremote-sensingsynthetic-datavision-language-models
- Vendors
- Hugging Face
- Products
- KnowChange
From the original article
1 min read at huggingface.co
Change data synthesis provides a cost-effective solution for expanding training data and improving the performance of change detection models.
By Yaoyi Qi, Xingxing Weng, Chao Pang · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.