Tackling Few-Shot Segmentation
in Remote Sensing
via Inpainting Diffusion Model

TelePIX
ICLR 2025 Machine Learning for Remote Sensing Workshop (Oral)
main_arch

Overall pipeline of the proposed approach.
An inpainting diffusion model generates novel-class samples, and SAM refines the segmentation masks. The results are used for training samples to improve model performance on few-shot settings.

Abstract

Limited data is a common problem in remote sensing due to the high cost of obtaining annotated samples. In the few-shot segmentation task, models are typically trained on base classes with abundant annotations and later adapted to novel classes with limited examples. However, this often necessitates specialized model architectures or complex training strategies. Instead, we propose a simple approach that leverages diffusion models to generate diverse variations of novel-class objects within a given scene, conditioned by the limited examples of the novel classes. By framing the problem as an image inpainting task, we synthesize plausible instances of novel classes under various environments, effectively increasing the number of samples for the novel classes and mitigating overfitting. The generated samples are then assessed using a cosine similarity metric to ensure semantic consistency with the novel classes. Additionally, we employ Segment Anything Model (SAM) to segment the generated samples and obtain precise annotations. By using high-quality synthetic data, we can directly fine-tune off-the-shelf segmentation models. Experimental results demonstrate that our method significantly enhances segmentation performance in low-data regimes, highlighting its potential for real-world remote sensing applications.

Qualitative Results

Quantitative Results

tabcompare

IoU comparison of various methods on different object classes.
The underline indicates best results in each group and the boldface indicates best results overall.

Citation

@article{2025rspaint,
  title={Tackling Few-Shot Segmentation in Remote Sensing via Inpainting Diffusion Model},
  author={Immanuel, Steve Andreas and Cho, Woojin and Heo, Junhyuk and Kwon, Darongsae},
  journal={arXiv preprint arXiv:2503.03785},
  year={2025}
}