Generating Satellite Imagery Data for Wildfire Detection through Mask-Conditioned Generative AI

Researchers developed a generative AI approach using EarthSynth to create synthetic post-wildfire satellite imagery for training deep learning wildfire detection systems. The study found that inpainting-based pipelines significantly outperformed full-tile generation, achieving better spatial alignment and burn area detection accuracy.

• →Scarcity of labeled satellite imagery remains a major bottleneck for AI-based wildfire monitoring systems.
• →EarthSynth diffusion model can synthesize realistic post-wildfire imagery without task-specific retraining.
• →Inpainting with pre-fire context consistently outperforms full-tile generation across all evaluation metrics.
• →VLM-assisted prompt generation is competitive with hand-crafted prompts for generating synthetic wildfire data.
• →The approach provides a foundation for incorporating generative data augmentation into wildfire detection pipelines.