Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ChangeAnywhere: Sample Generation for Remote Sensing Change Detection via Semantic Latent Diffusion Model

Kai Tang, Jin Chen

TL;DR

This work tackles the labeling burden in remote sensing change detection by introducing ChangeAnywhere, a diffusion-based pipeline that generates bi-temporal, semantically annotated CD data from large single-temporal semantic datasets. A semantic latent diffusion model is trained in latent space on OpenEarthMap masks and conditioned to produce consistent bi-temporal pairs, with change events simulated through semantic mask manipulation and latent-space mixing. The method yields Ce-100K, the largest synthetic CD dataset, and pre-training CD models on Ce-100K delivers notable zero-shot and few-shot gains on benchmark datasets, validating the approach's value for CD and foundation-model development. By enabling diverse, semantically constrained CD data without multi-temporal labeling, ChangeAnywhere offers a practical path toward scalable remote sensing CD pre-training and improved model generalization.

Abstract

Remote sensing change detection (CD) is a pivotal technique that pinpoints changes on a global scale based on multi-temporal images. With the recent expansion of deep learning, supervised deep learning-based CD models have shown satisfactory performance. However, CD sample labeling is very time-consuming as it is densely labeled and requires expert knowledge. To alleviate this problem, we introduce ChangeAnywhere, a novel CD sample generation method using the semantic latent diffusion model and single-temporal images. Specifically, ChangeAnywhere leverages the relative ease of acquiring large single-temporal semantic datasets to generate large-scale, diverse, and semantically annotated bi-temporal CD datasets. ChangeAnywhere captures the two essentials of CD samples, i.e., change implies semantically different, and non-change implies reasonable change under the same semantic constraints. We generated ChangeAnywhere-100K, the largest synthesis CD dataset with 100,000 pairs of CD samples based on the proposed method. The ChangeAnywhere-100K significantly improved both zero-shot and few-shot performance on two CD benchmark datasets for various deep learning-based CD models, as demonstrated by transfer experiments. This paper delineates the enormous potential of ChangeAnywhere for CD sample generation and demonstrates the subsequent enhancement of model performance. Therefore, ChangeAnywhere offers a potent tool for remote sensing CD. All codes and pre-trained models will be available at https://github.com/tangkai-RS/ChangeAnywhere.

ChangeAnywhere: Sample Generation for Remote Sensing Change Detection via Semantic Latent Diffusion Model

TL;DR

This work tackles the labeling burden in remote sensing change detection by introducing ChangeAnywhere, a diffusion-based pipeline that generates bi-temporal, semantically annotated CD data from large single-temporal semantic datasets. A semantic latent diffusion model is trained in latent space on OpenEarthMap masks and conditioned to produce consistent bi-temporal pairs, with change events simulated through semantic mask manipulation and latent-space mixing. The method yields Ce-100K, the largest synthetic CD dataset, and pre-training CD models on Ce-100K delivers notable zero-shot and few-shot gains on benchmark datasets, validating the approach's value for CD and foundation-model development. By enabling diverse, semantically constrained CD data without multi-temporal labeling, ChangeAnywhere offers a practical path toward scalable remote sensing CD pre-training and improved model generalization.

Abstract

Remote sensing change detection (CD) is a pivotal technique that pinpoints changes on a global scale based on multi-temporal images. With the recent expansion of deep learning, supervised deep learning-based CD models have shown satisfactory performance. However, CD sample labeling is very time-consuming as it is densely labeled and requires expert knowledge. To alleviate this problem, we introduce ChangeAnywhere, a novel CD sample generation method using the semantic latent diffusion model and single-temporal images. Specifically, ChangeAnywhere leverages the relative ease of acquiring large single-temporal semantic datasets to generate large-scale, diverse, and semantically annotated bi-temporal CD datasets. ChangeAnywhere captures the two essentials of CD samples, i.e., change implies semantically different, and non-change implies reasonable change under the same semantic constraints. We generated ChangeAnywhere-100K, the largest synthesis CD dataset with 100,000 pairs of CD samples based on the proposed method. The ChangeAnywhere-100K significantly improved both zero-shot and few-shot performance on two CD benchmark datasets for various deep learning-based CD models, as demonstrated by transfer experiments. This paper delineates the enormous potential of ChangeAnywhere for CD sample generation and demonstrates the subsequent enhancement of model performance. Therefore, ChangeAnywhere offers a potent tool for remote sensing CD. All codes and pre-trained models will be available at https://github.com/tangkai-RS/ChangeAnywhere.
Paper Structure (11 sections, 6 equations, 3 figures, 5 tables)

This paper contains 11 sections, 6 equations, 3 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Preliminaries
  4. Semantic Latent Diffusion Model
  5. Change Events Simulation
  6. Change Sample Generation
  7. Experiments
  8. Experimental Setting
  9. Visualization for Ce-100K
  10. Conclusion
  11. Acknowledgements

Figures (3)

  • Figure 1: The overall flow of ChangeAnywhere. (a) Semantic latent diffusion model training. (b) Change events simulation. (c) Change Sample Generation. All the diffusion and denoising processes are performed in the latent space, and we omit the process of encoding and decoding the original remote sensing images using VQGAN.
  • Figure 2: Various change detect models performance curves on the SECOND dataset.
  • Figure 3: Visualization of ChangeAnywhere-100K. Legend: Tree, Bare land, Rangeland, Developed space, Water, Agriculture land, Building, and white replaces Road.