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

2026-04-02

AI Summary

This study addresses the scarcity of labeled satellite imagery for wildfire monitoring by introducing, for the first time, the mask-conditional diffusion model EarthSynth to synthesize post-disaster Sentinel-2 RGB images. Leveraging an image inpainting architecture combined with structured prompting strategies—including handwritten cues and vision-language prompts generated by Qwen2-VL—the method enables controllable generation without task-specific fine-tuning. A region-aware color-matching post-processing step further enhances photorealism. Experimental results demonstrate that the inpainting paradigm significantly outperforms full-image generation across four key metrics: Burn IoU (peaking at 0.456), color distance, darkness contrast, and spectral plausibility, offering an effective data augmentation solution for wildfire detection.

Abstract

The scarcity of labeled satellite imagery remains a fundamental bottleneck for deep-learning (DL)-based wildfire monitoring systems. This paper investigates whether a diffusion-based foundation model for Earth Observation (EO), EarthSynth, can synthesize realistic post-wildfire Sentinel-2 RGB imagery conditioned on existing burn masks, without task-specific retraining. Using burn masks derived from the CalFireSeg-50 dataset (Martin et al., 2025), we design and evaluate six controlled experimental configurations that systematically vary: (i) pipeline architecture (mask-only full generation vs. inpainting with pre-fire context), (ii) prompt engineering strategy (three hand-crafted prompts and a VLM-generated prompt via Qwen2-VL), and (iii) a region-wise color-matching post-processing step. Quantitative assessment on 10 stratified test samples uses four complementary metrics: Burn IoU, burn-region color distance (ΔC_burn), Darkness Contrast, and Spectral Plausibility. Results show that inpainting-based pipelines consistently outperform full-tile generation across all metrics, with the structured inpainting prompt achieving the best spatial alignment (Burn IoU = 0.456) and burn saliency (Darkness Contrast = 20.44), while color matching produces the lowest color distance (ΔC_burn = 63.22) at the cost of reduced burn saliency. VLM-assisted inpainting is competitive with hand-crafted prompts. These findings provide a foundation for incorporating generative data augmentation into wildfire detection pipelines. Code and experiments are available at: https://www.kaggle.com/code/valeriamartinh/genai-all-runned

Problem

• wildfire detection
• satellite imagery
• data scarcity
• labeled data
• Earth Observation

Innovation

• mask-conditioned generation
• diffusion model
• wildfire detection
• satellite imagery synthesis
• inpainting