Leveraging generative AI for cross-regional small object detection in satellite imagery

Zheyang Qin, Stanislav Panev, Celso M de Melo, Shayok Chakraborty, Jessica K. Hodgins, Fernando De la Torre

Abstract

Detecting small objects like cars in satellite imagery is challenging, especially when labeled data is unavailable for a target region. This study introduces a novel approach using generative AI to address this issue by fine-tuning a Stable Diffusion model on images from one geographical region, the source (Selwyn, New Zealand), and a second region of interest, the target (Utah, USA) environment. Our framework employs state-of-the-art cross and self-attention mechanisms alongside the CLIPSeg image segmentation method to generate high-quality synthetic datasets with minimal supervision. Our empirical results show a significant improvement in detection performance, improving the accuracy by over 20% when testing the target dataset compared to the performance of a baseline model trained only on source data.
BibTeX
@inproceedings{qin2026leveraging,
  author = {Zheyang Qin and Stanislav Panev and Celso M de Melo and Shayok Chakraborty and Jessica K. Hodgins and Fernando De la Torre},
  title = {{Leveraging generative AI for cross-regional small object detection in satellite imagery}},
  booktitle = {SPIE 13459, Synthetic Data for Artificial Intelligence and Machine Learning: Tools, Techniques, and Applications III},
  year = {2026},
  url = {https://openreview.net/forum?id=4ZcklY5oB6},
}

Loading related content...

Figure