Make Some Noise: Unsupervised Remote Sensing Change Detection Using Latent Space Perturbations

TL;DR

MaSoN tackles unsupervised remote-sensing change detection by training on changes synthesized directly in the latent feature space. It injects two decoupled Gaussian noises—irrelevant and relevant—whose scales are dynamically estimated from target data statistics, enabling diverse, data-aligned changes without external labels. The method achieves state-of-the-art results across five benchmarks (average F1 improvement of 14.1 percentage points) and generalises to multispectral and SAR modalities through a simple encoder swap. These results underscore latent-space perturbations as a powerful, data-efficient approach for robust change detection in varied remote-sensing scenes.

Abstract

Unsupervised change detection (UCD) in remote sensing aims to localise semantic changes between two images of the same region without relying on labelled data during training. Most recent approaches rely either on frozen foundation models in a training-free manner or on training with synthetic changes generated in pixel space. Both strategies inherently rely on predefined assumptions about change types, typically introduced through handcrafted rules, external datasets, or auxiliary generative models. Due to these assumptions, such methods fail to generalise beyond a few change types, limiting their real-world usage, especially in rare or complex scenarios. To address this, we propose MaSoN (Make Some Noise), an end-to-end UCD framework that synthesises diverse changes directly in the latent feature space during training. It generates changes that are dynamically estimated using feature statistics of target data, enabling diverse yet data-driven variation aligned with the target domain. It also easily extends to new modalities, such as SAR. MaSoN generalises strongly across diverse change types and achieves state-of-the-art performance on five benchmarks, improving the average F1 score by 14.1 percentage points. Project page: https://blaz-r.github.io/mason_ucd

Figures (18)

• Figure 1: Related methods that rely on foundation models in a training-free manner or on changes generated in pixel-space often produce noisy or inaccurate masks, either missing relevant changes or overreacting to irrelevant changes. By generating changes in latent space, MaSoN learns from diverse and more data-aligned variations, producing better predictions with fewer false positives. This highlights MaSoN's improved ability to generalise across diverse and challenging change scenarios.
• Figure 1: Distributions of feature differences $f_1^{(l)} - f_2^{(l)}$, on the SYSU Dataset.
• Figure 2: Histogram plot of feature differences $f_1^{(l)} - f_2^{(l)}$, averaged across all five datasets and all channels per layer. Unchanged regions are narrowly concentrated near zero, while changed regions exhibit broader variation, especially in deeper layers. Both distributions can be approximated by a zero-centred Gaussian, but each with a different variance parameter. This directly motivates our latent-space change generation strategy.
• Figure 2: Distributions of feature differences $f_1^{(l)} - f_2^{(l)}$, on the LEVIR Dataset.
• Figure 3: The architecture of the proposed unsupervised change detection framework, MaSoN.