Deep generative networks have recently achieved unprecedented performance in precise image and video editing using sophisticated textual prompts. However, the effectiveness of such models heavily depends on access to very large supervised and annotated image datasets, which can be very difficult to obtain. This is particularly true for satellite instruments, which very rarely overlap with labelled data, and suffer from domain shifts in the rare occasions they do. In this paper, we investigate the potential of flow matching models for unsupervised domain adaptation of satellite radiometer images. Our main contribution is a novel unsupervised method that achieves precise domain alignment by leveraging parts of the deterministic ordinary differential equations in flow matching models, conditioned on different satellite instruments. A key strength of our approach is its ability to preserve essential information while adapting across any domains since the perturbations are in theory bijective. Extensive experiments conducted on the GPM-Core constellation show the benefit of our conditional domain adaptation, particularly in improving rain precipitation estimation from radiometer imagery.
Figures & tables
Figure 1 : Overview of the proposed conditional flow matching framework. During training (left), the model learns to map both GMI and F18 images to a standard Gaussian distribution using domain-specific labels. During inference (right), domain adaptation is performed in two steps: first, an F18 image is noised by running the forward ODE from t0 to t1 with the associated label conditioning ( yi in Eq. 8 ). Afterwards, the reverse ODE is solved from t1 to t0 using the GMI conditioning parameters instead ( yj in Eq. 8 ), so as to obtain the adapted image in the GMI domain (this corresponds to xˉi→j in Eq. 8 ).
Instrument
Central Frequency
Polarisation
Bandwidth
Pixel Size (km)
GMI
36.6GHz
V, H
1000 MHz
6.0×13.4
SSMI/S
37.0GHz
V, H
1580 MHz
25.0×12.5
GMI
89.0GHz
V, H
6000 MHz
3.0×13.4
SSMI/S
91.6GHz
V, H
2829 MHz
12.5×12.5
Table 1 : Technical information on the considered GMI and SSMI/S, grouped by frequency bands and polarisations (V vertical, H horizontal).
RR in mm/h
Model
RMSE
MAE
ME
[0.0,0.1)
F18 Baseline
0.074
0.011
0.011
SDEdit [ 13 ]
0.147
0.019
0.019
DDIB [ 19 ]
0.102
0.012
0.012
CycleGAN [ 16 ]
0.101
0.016
0.016
Ours
0.071
0.007
0.007
[0.1,2.5)
F18 Baseline
0.638
0.412
-0.267
Table 2 : RMSE, MAE, and ME for different rain rate intervals.
Figure 2 : Comparison of precipitation estimates (mm/h) obtained using DRAIN on the ground-truth GMI ( 2(a) ), the unadapted F18 baseline ( 2(b) ), our method ( 2(c) ), and the comparative methods DDIB [ 19 ] , SDEdit [ 13 ] , and CycleGAN [ 16 ] ( 2(d) , 2(e) , 2(f) ). Results are shown for the 2019-02-06 overpass all projected on GMI pixels. Note that perfect collocation is unattainable due to differences in observation angles, acquisition times, and pixel counts, which may cause discrepancies despite “perfect” adaptation.
Figure 3 : Comparison between the original F18 images (37GHz-V channel) ( 3(a) ), with the adapted images generated using our proposed method ( 3(b) ), as well as the comparative methods DDIB [ 19 ] ( 3(c) ), SDEdit [ 13 ] ( 3(d) ), and CycleGAN [ 16 ] ( 3(e) ).
Generative machine learning is an increasingly important complement to dynamical downscaling for producing high-resolution precipitation projections, with diffusion models currently the leading approach. Flow matching is a related generative framework that has recently achieved strong results across image, video and other domains, and shown early promise for downscaling. We train a flow matching model to map daily precipitation from 8 km to 2 km over a convective-scale domain centred on Singapore, and benchmark it against CPMGEM, a score-based diffusion model. Flow matching achieves consistently better spatial skill: higher fractions skill score at every precipitation threshold and neighbourhood scale tested, and tighter structure and amplitude components of the SAL score with comparable location skill. However, flow matching underestimates the upper tail of the precipitation distribution, resulting in a dry bias in the climatological mean. These results suggest that flow matching is a competitive generative framework for convective-scale precipitation downscaling, particularly well suited to capturing spatial structure.
Developing robust flood assessment models requires high-quality paired satellite imagery, yet such data remain scarce for flood-specific image generation. Although generative models provide a promising means of data augmentation, existing methods often yield implausible spatial layouts of flooded regions and distort scene structures. We propose FlowForm, a framework for satellite flood synthesis that integrates SWE-inspired latent regularization with structure-aware conditioning. The Flood Descriptor Module (FDM) imposes differentiable penalties on residuals of the steady-state Shallow Water Equation in auxiliary latent fields at the diffusion bottleneck. The Terrain Anchor Adapter (TAA) injects depth, semantic, and edge features at four encoder scales of the U-Net. We further curate FloodScape, a large-scale, high-resolution dataset comprising paired satellite images acquired before and after disasters. In addition to standard image-generation metrics, we evaluate the consistency of flooded regions, zero-shot generalization to a geographically held-out flood event, and sensitivity to individual components. Across all reported comparisons, FlowForm achieves higher visual fidelity, greater similarity between paired images, and stronger consistency of flooded regions.
Unsupervised domain adaptation (UDA) for semantic segmentation remains challenging under adverse weather conditions because severe appearance changes enlarge the domain gap and degrade the reliability of pseudo labels in the target domain. To address this problem, we propose an Intra-Class Mixing Consistency (ICM) framework that enforces prediction consistency between an intra-class mixed image and its original counterpart. Unlike previous mixing-based consistency methods that combine regions across different images or domains and may introduce unrealistic semantic inconsistencies, ICM performs mixing within the same image and semantic class, preserving realistic semantic layout for consistency regularization. With ICM, we establish a new state-of-the-art performance for clear-to-adverse-weather unsupervised domain adaptation (UDA) in semantic segmentation. On the Cityscapes → ACDC benchmark, our method achieves 75.7% mIoU, outperforming the previous state of the art by +1.9 pp, demonstrating its effectiveness in mitigating class confusion under challenging environmental conditions. The code is provided in the supplementary material.