ChangeDiff: A Multi-Temporal Change Detection Data Generator with Flexible Text Prompts via Diffusion Model

Qi Zang, Jiayi Yang, Shuang Wang, Dong Zhao, Wenjun Yi, Zhun Zhong

Abstract:

progress in change detection (CD) with the support of pixel
level annotations. However, collecting diverse data and man
ually annotating them is costly, laborious, and knowledge
intensive. Existing generative methods for CD data synthe
sis show competitive potential in addressing this issue but
still face the following limitations: 1) difficulty in flexibly
controlling change events, 2) dependence on additional data
to train the data generators, 3) focus on specific change de
tection tasks. To this end, this paper focuses on the seman
tic CD (SCD) task and develops a multi-temporal SCD data
generator ChangeDiff by exploring powerful diffusion mod
els. ChangeDiff innovatively generates change data in two
steps: first, it uses text prompts and a text-to-layout (T2L)
model to create continuous layouts, and then it employs
layout-to-image (L2I) to convert these layouts into images.
Specifically, we propose multi-class distribution-guided text
prompts (MCDG-TP), allowing for layouts to be generated
f
lexibly through controllable classes and their correspond
ing ratios. Subsequently, to generalize the T2L model to the
proposed MCDG-TP, a class distribution refinement loss is
further designed as training supervision. Our generated data
shows significant progress in temporal continuity, spatial di
versity, and quality realism, empowering change detectors
with accuracy and transferability.