pith. sign in

arxiv: 2507.04678 · v3 · submitted 2025-07-07 · 💻 cs.CV

ChangeBridge: Spatiotemporal Image Generation with Multimodal Controls for Remote Sensing

Zhenghui Zhao , Chen Wu , Xiangyong Cao , Di Wang , Hongruixuan Chen , Datao Tang , Liangpei Zhang , Zhuo Zheng This is my paper

Pith reviewed 2026-05-19 06:53 UTC · model grok-4.3

classification 💻 cs.CV
keywords spatiotemporal generationremote sensingdiffusion bridgechange detectionmultimodal controldrift mapimage synthesis
0
0 comments X

The pith

ChangeBridge generates post-event remote sensing scenes that are spatially and temporally coherent using a drift-asynchronous diffusion bridge.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents ChangeBridge as a model for generating future remote sensing images that handle both specific events such as new constructions and broader temporal variations like seasonal changes. Current approaches struggle because they focus only on event-driven alterations and overlook how time passes between observations. ChangeBridge addresses this by initializing the diffusion process from a composed pre-event state, applying a pixel-wise drift map to differentiate the strength of changes for events versus ongoing evolution, and using a drift-aware network for reconstruction. If this works, it enables more realistic simulations of landscape changes over time, supporting applications in urban planning and creating synthetic data for training change detection systems.

Core claim

The central claim is that a drift-asynchronous diffusion bridge, built from composed bridge initialization, asynchronous drift diffusion, and drift-aware denoising, allows generation of post-event scenes conditioned on pre-event images and multimodal event controls that maintain both spatial alignment and temporal coherence.

What carries the argument

A drift-asynchronous diffusion bridge that uses a pixel-wise drift map to assign different drift magnitudes to event-driven changes and cross-temporal variations during the pre-to-post transition.

If this is right

  • ChangeBridge produces better cross-spatiotemporal aligned scenarios than existing methods.
  • It has potential applications in land-use planning.
  • It can act as a data generation engine for various change detection tasks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This separation of change types could help in modeling complex scenarios like climate-induced shifts combined with human activities.
  • Similar drift mechanisms might improve generation tasks in other time-series imaging domains such as medical scans.
  • Integrating more control modalities could further enhance precision in predicting specific outcomes.

Load-bearing premise

Multimodal event controls together with the pixel-wise drift map separate event-driven changes from cross-temporal variations without introducing uncorrectable artifacts in the denoising process.

What would settle it

Comparing generated images against actual post-event observations and finding systematic mismatches in alignment or introduced artifacts in areas of mixed change types.

Figures

Figures reproduced from arXiv: 2507.04678 by Chen Wu, Datao Tang, Di Wang, Hongruixuan Chen, Liangpei Zhang, Xiangyong Cao, Zhenghui Zhao, Zhuo Zheng.

Figure 1
Figure 1. Figure 1: ChangeBridge generates post-event images from pre-event inputs, conditioned on [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Comparison between (a) typical multi-conditional dif [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Framework of ChangeBridge, which construct a conditional spatiotemporal diffusion bridge process for cross-spatiotemporal [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison to methods conditioned on instance lay￾outs. These methods of image editing include UNITE, and Con￾trolNet incorporating IP-Adapter (IPA) [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Comparison to methods conditioned on [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Comparison to methods conditioned on text prompts. These methods of image editing include DreamBooth and ELITE. first row shows results of real data, serving as the upper bound. For semantic map-conditioned generation, Change￾Bridge outperforms existing methods on the SECOND dataset, achieving the lowest FID of 98.42 and the high￾est IS of 5.57. Compared to UNITE and ControlNet+IPA, which yield FID scores … view at source ↗
Figure 7
Figure 7. Figure 7: Example Diversity of our ChangeBridge, where sam [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
read the original abstract

Spatiotemporal image generation is a highly meaningful task, which can generate future scenes conditioned on given observations. However, existing change generation methods can only handle event-driven changes (e.g., new buildings) and fail to model cross-temporal variations (e.g., seasonal shifts). In this work, we propose ChangeBridge, a conditional spatiotemporal image generation model for remote sensing. Given pre-event images and multimodal event controls, ChangeBridge generates post-event scenes that are both spatially and temporally coherent. The core idea is a drift-asynchronous diffusion bridge. Specifically, it consists of three main modules: a) Composed Bridge Initialization, which replaces noise initialization. It starts the diffusion from a composed pre-event state, modeling a diffusion bridge process. b) Asynchronous Drift Diffusion, which uses a pixel-wise drift map, assigning different drift magnitudes to event and temporal evolution. This enables differentiated generation during the pre-to-post transition. c) Drift-Aware Denoising, which embeds the drift map into the denoising network, guiding drift-aware reconstruction. Experiments show that ChangeBridge can generate better cross-spatiotemporal aligned scenarios compared to state-of-the-art methods. Additionally, ChangeBridge shows great potential for land-use planning and as a data generation engine for a series of change detection tasks. Code is available at https://github.com/zhenghuizhao/ChangeBridge

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces ChangeBridge, a conditional diffusion model for spatiotemporal image generation in remote sensing. Given pre-event images and multimodal event controls, it generates post-event scenes via a drift-asynchronous diffusion bridge comprising composed bridge initialization (replacing standard noise), asynchronous drift diffusion (using a pixel-wise drift map to assign differentiated magnitudes to event-driven changes versus cross-temporal variations), and drift-aware denoising (embedding the map into the network). The central claim is that this produces better cross-spatiotemporal alignment than existing methods, with applications to land-use planning and synthetic data for change detection; code is released.

Significance. If the drift map successfully isolates asynchronous changes without uncorrectable residuals, the approach could meaningfully extend conditional diffusion to remote-sensing scenarios that mix abrupt events and gradual temporal shifts, improving upon methods limited to event-driven changes only. Open-sourced code is a clear strength for reproducibility.

major comments (2)
  1. [Abstract / Experiments] Abstract and experimental section: the claim of superior cross-spatiotemporal alignment is stated without any reported quantitative metrics (e.g., FID, PSNR, change-detection F1 on generated pairs), ablation studies, or baseline details. This leaves the headline result unverified from the provided text.
  2. [Asynchronous Drift Diffusion] Asynchronous Drift Diffusion module: the pixel-wise drift map is presented as the mechanism that separates event-driven changes from temporal evolution, yet no ablation (e.g., uniform-drift control) or direct validation (per-pixel fidelity to ground-truth change masks, or downstream change-detection accuracy on generated pairs) is described to confirm separation succeeds rather than merely amplifying conditioning strength.
minor comments (1)
  1. [Composed Bridge Initialization] Clarify the precise construction of the composed pre-event state in the initialization module and how multimodal controls are encoded before being fed to the drift map.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the major comments point by point below, indicating where revisions will be made to improve the clarity and verifiability of our results.

read point-by-point responses

  1. Referee: [Abstract / Experiments] Abstract and experimental section: the claim of superior cross-spatiotemporal alignment is stated without any reported quantitative metrics (e.g., FID, PSNR, change-detection F1 on generated pairs), ablation studies, or baseline details. This leaves the headline result unverified from the provided text.

    Authors: The current manuscript emphasizes qualitative visual comparisons in the experiments to illustrate the advantages in cross-spatiotemporal coherence. However, we agree that quantitative support would strengthen the claims. In the revised version, we will report standard metrics such as FID and PSNR for image quality, as well as change-detection F1 scores on pairs generated by ChangeBridge versus baselines. We will also expand the experimental section with explicit baseline details and additional ablations. revision: yes

  2. Referee: [Asynchronous Drift Diffusion] Asynchronous Drift Diffusion module: the pixel-wise drift map is presented as the mechanism that separates event-driven changes from temporal evolution, yet no ablation (e.g., uniform-drift control) or direct validation (per-pixel fidelity to ground-truth change masks, or downstream change-detection accuracy on generated pairs) is described to confirm separation succeeds rather than merely amplifying conditioning strength.

    Authors: We recognize that additional evidence is required to validate the specific contribution of the pixel-wise drift map. To address this, we will add an ablation experiment using a uniform drift map as a control. Furthermore, we will include quantitative validation by measuring per-pixel agreement with ground-truth change masks and assessing the accuracy of change detection models trained on the synthetically generated data. These additions will help demonstrate that the asynchronous drift mechanism provides meaningful separation beyond enhanced conditioning. revision: yes

Circularity Check

0 steps flagged

No significant circularity in architectural derivation or claims

full rationale

The paper presents ChangeBridge as a new conditional diffusion model with three explicitly described modules (Composed Bridge Initialization replacing noise, Asynchronous Drift Diffusion via a pixel-wise drift map derived from multimodal controls and pre-event images, and Drift-Aware Denoising). These are introduced as independent design choices rather than derived quantities. No equations are shown that reduce outputs to inputs by construction, no parameters are fitted to a data subset and then relabeled as predictions, and no uniqueness theorems or ansatzes are imported via self-citation to force the central construction. The abstract's claim of better cross-spatiotemporal alignment is positioned as an experimental outcome, not a definitional necessity. This matches the default expectation for most papers and yields no load-bearing circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the assumption that a diffusion bridge initialized from composed pre-event states plus a learned pixel-wise drift map can separate event and temporal evolution without additional supervision. No free parameters are explicitly named in the abstract. No new physical entities are postulated.

axioms (1)
  • domain assumption Multimodal controls can be injected into the diffusion process to guide both spatial and temporal coherence.
    Invoked in the description of the overall conditioning mechanism.
invented entities (1)
  • pixel-wise drift map no independent evidence
    purpose: Assign different drift magnitudes to event-driven versus temporal changes during the pre-to-post transition.
    Introduced as part of the Asynchronous Drift Diffusion module; no independent evidence outside the model is provided.

pith-pipeline@v0.9.0 · 5796 in / 1252 out tokens · 15458 ms · 2026-05-19T06:53:25.098779+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

49 extracted references · 49 canonical work pages · 1 internal anchor

  1. [1]

    Satdm: Synthe- sizing realistic satellite image with semantic layout conditioning using diffusion models

    Orkhan Baghirli, Hamid Askarov, Imran Ibrahimli, Is- mat Bakhishov, and Nabi Nabiyev. Satdm: Synthe- sizing realistic satellite image with semantic layout conditioning using diffusion models. arXiv preprint arXiv:2309.16812, 2023. 1

    work page arXiv 2023

  2. [2]

    A transformer-based siamese network for change de- tection

    Wele Gedara Chaminda Bandara and Vishal M Patel. A transformer-based siamese network for change de- tection. In IGARSS 2022-2022 IEEE International Geoscience and Remote Sensing Symposium , pages 207–210. IEEE, 2022. 7

    work page 2022

  3. [3]

    Bourdis, D

    N. Bourdis, D. Marraud, and H. Sahbi. Constrained optical flow for aerial image change detection. In2011 IEEE International Geoscience and Remote Sensing Symposium, pages 4176–4179. IEEE, 2011. 3

    work page 2011

  4. [4]

    Machine learning al- gorithms for urban land use planning: A review

    V Chaturvedi and WT de Vries. Machine learning al- gorithms for urban land use planning: A review. Ur- ban Science, 5(3):68, 2021. 3

    work page 2021

  5. [5]

    Spectral-cascaded diffu- sion model for remote sensing image spectral super- resolution

    Bowen Chen, Liqin Liu, Chenyang Liu, Zhengxia Zou, and Zhenwei Shi. Spectral-cascaded diffu- sion model for remote sensing image spectral super- resolution. IEEE Transactions on Geoscience and Re- mote Sensing, 2024. 1

    work page 2024

  6. [6]

    A spatial-temporal attention-based method and a new dataset for remote sensing image change detection

    Hao Chen and Zhenwei Shi. A spatial-temporal attention-based method and a new dataset for remote sensing image change detection. Remote Sensing, 12 (10):1662, 2020. 5, 7

    work page 2020

  7. [7]

    Remote sens- ing image change detection with transformers

    Hao Chen, Zipeng Qi, and Zhenwei Shi. Remote sens- ing image change detection with transformers. IEEE Transactions on Geoscience and Remote Sensing, 60: 1–14, 2021. 7

    work page 2021

  8. [8]

    A rule-based model of urban land use and economic growth

    A T Crooks, N W Gibin, and J J Heppenstall. A rule-based model of urban land use and economic growth. International Journal of Geographical Infor- mation Science, 24(2):311–331, 2010. 3

    work page 2010

  9. [9]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances in neural information processing systems , 34:8780– 8794, 2021. 2

    work page 2021

  10. [10]

    Taming transformers for high-resolution image syn- thesis

    Patrick Esser, Robin Rombach, and Bjorn Ommer. Taming transformers for high-resolution image syn- thesis. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 12873–12883, 2021. 6 9

    work page 2021

  11. [11]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium

    Martin Heusel, Hubert Ramsauer, Thomas Un- terthiner, Bernhard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilibrium. Advances in neural infor- mation processing systems, 30, 2017. 6

    work page 2017

  12. [12]

    Denoising diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. In NeurIPS, 2020. 2, 5

    work page 2020

  13. [13]

    Resolving multi-condition confusion for finetuning-free personalized image gen- eration

    Qihan Huang, Siming Fu, Jinlong Liu, Hao Jiang, Yipeng Yu, and Jie Song. Resolving multi-condition confusion for finetuning-free personalized image gen- eration. arXiv preprint arXiv:2409.17920, 2024. 5

    work page arXiv 2024

  14. [14]

    Fully con- volutional networks for multisource building extrac- tion from an open aerial and satellite imagery data set

    Shunping Ji, Shiqing Wei, and Meng Lu. Fully con- volutional networks for multisource building extrac- tion from an open aerial and satellite imagery data set. IEEE Transactions on Geoscience and Remote Sens- ing, 57(1):574–586, 2018. 5

    work page 2018

  15. [15]

    Diffusionsat: A generative foundation model for satellite imagery.arXiv preprint arXiv:2312.03606, 2023

    Samar Khanna, Patrick Liu, Linqi Zhou, Chenlin Meng, Robin Rombach, Marshall Burke, David Lo- bell, and Stefano Ermon. Diffusionsat: A generative foundation model for satellite imagery. arXiv preprint arXiv:2312.03606, 2023. 1

    work page arXiv 2023

  16. [16]

    Bbdm: Image-to-image translation with brownian bridge dif- fusion models

    Bo Li, Kaitao Xue, Bin Liu, and Yu-Kun Lai. Bbdm: Image-to-image translation with brownian bridge dif- fusion models. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern Recognition , pages 1952–1961, 2023. 2, 3

    work page 1952

  17. [17]

    Open-cd: A comprehensive toolbox for change detection

    Kaiyu Li, Jiawei Jiang, Andrea Codegoni, Chengxi Han, Yupeng Deng, Keyan Chen, Zhuo Zheng, Hao Chen, Zhengxia Zou, Zhenwei Shi, et al. Open-cd: A comprehensive toolbox for change detection. arXiv preprint arXiv:2407.15317, 2024. 7

    work page arXiv 2024

  18. [18]

    Gligen: Open-set grounded text-to- image generation

    Yuheng Li, Haotian Liu, Qingyang Wu, Fangzhou Mu, Jianwei Yang, Jianfeng Gao, Chunyuan Li, and Yong Jae Lee. Gligen: Open-set grounded text-to- image generation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recogni- tion, pages 22511–22521, 2023. 2

    work page 2023

  19. [19]

    Diverse hyperspectral remote sens- ing image synthesis with diffusion models

    Liqin Liu, Bowen Chen, Hao Chen, Zhengxia Zou, and Zhenwei Shi. Diverse hyperspectral remote sens- ing image synthesis with diffusion models. IEEE Transactions on Geoscience and Remote Sensing, 61: 1–16, 2023. 1

    work page 2023

  20. [20]

    A diffusion pro- cess and its applications to detecting a change in the drift of brownian motion

    Moshe Pollak and David Siegmund. A diffusion pro- cess and its applications to detecting a change in the drift of brownian motion. Biometrika, 72(2):267–280,

    work page

  21. [21]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sas- try, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning, pages 8748–8763. PMLR, 2021. 6

    work page 2021

  22. [22]

    Spatially explicit simulation of land use/land cover changes: Current coverage and future prospects

    Y Ren, Y L ¨u, A Comber, B Fu, P Harris, and L Wu. Spatially explicit simulation of land use/land cover changes: Current coverage and future prospects. Land Use Policy, 80:324–336, 2019. 3

    work page 2019

  23. [23]

    High- resolution image synthesis with latent diffusion mod- els

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High- resolution image synthesis with latent diffusion mod- els. In CVPR, 2022. 2, 3, 6

    work page 2022

  24. [24]

    A novel al- gorithm for calculating transition potential in cellular automata models of land-use/cover change

    MS Roodposhti, J Aryal, and BA Bryan. A novel al- gorithm for calculating transition potential in cellular automata models of land-use/cover change. Science of the Total Environment, 674:290–303, 2019. 3

    work page 2019

  25. [25]

    Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Yael Pritch, Michael Rubinstein, and Kfir Aberman. Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 22500–22510, 2023. 2, 6

    work page 2023

  26. [26]

    Improved techniques for training gans

    Tim Salimans, Ian Goodfellow, Wojciech Zaremba, Vicki Cheung, Alec Radford, and Xi Chen. Improved techniques for training gans. Advances in neural in- formation processing systems, 29, 2016. 6

    work page 2016

  27. [27]

    Unit-ddpm: Unpaired image translation with denoising diffusion probabilistic models,

    Hiroshi Sasaki, Chris G Willcocks, and Toby P Breckon. Unit-ddpm: Unpaired image translation with denoising diffusion probabilistic models. arXiv preprint arXiv:2104.05358, 2021. 2

    work page arXiv 2021

  28. [28]

    S2looking: A satellite side-looking dataset for build- ing change detection

    Li Shen, Yao Lu, Hao Chen, Hao Wei, Donghai Xie, Jiabao Yue, Rui Chen, Shouye Lv, and Bitao Jiang. S2looking: A satellite side-looking dataset for build- ing change detection. Remote Sensing, 13(24):5094,

    work page

  29. [29]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. In ICLR, 2021. 2

    work page 2021

  30. [30]

    Score-based generative modeling through stochastic differential equations

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. In ICLR, 2021. 2, 5

    work page 2021

  31. [31]

    Crs-diff: Con- trollable generative remote sensing foundation model

    Datao Tang, Xiangyong Cao, Xingsong Hou, Zhongyuan Jiang, and Deyu Meng. Crs-diff: Con- trollable generative remote sensing foundation model. arXiv e-prints, pages arXiv–2403, 2024. 1

    work page 2024

  32. [32]

    Changeanywhere: Sam- ple generation for remote sensing change detection via semantic latent diffusion model

    Kai Tang and Jin Chen. Changeanywhere: Sam- ple generation for remote sensing change detection via semantic latent diffusion model. arXiv preprint arXiv:2404.08892, 2024. 3

    work page arXiv 2024

  33. [33]

    A statistical model for land use and land cover change prediction: A case study of urban growth in the beijing metropolitan area

    X Tong, H Liu, and Y Li. A statistical model for land use and land cover change prediction: A case study of urban growth in the beijing metropolitan area. Envi- ronmental Modeling & Software , 153:105418, 2022. 3 10

    work page 2022

  34. [34]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. Advances in neural information processing systems , 30, 2017. 5

    work page 2017

  35. [35]

    On the theory of the brownian motion ii

    Ming Chen Wang and George Eugene Uhlenbeck. On the theory of the brownian motion ii. Reviews of mod- ern physics, 17(2-3):323, 1945. 3

    work page 1945

  36. [36]

    Elite: Encod- ing visual concepts into textual embeddings for cus- tomized text-to-image generation

    Yuxiang Wei, Yabo Zhang, Zhilong Ji, Jinfeng Bai, Lei Zhang, and Wangmeng Zuo. Elite: Encod- ing visual concepts into textual embeddings for cus- tomized text-to-image generation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 15943–15953, 2023. 6

    work page 2023

  37. [37]

    arXiv preprint arXiv:2010.05687 (2020)

    Kunping Yang, Gui-Song Xia, Zicheng Liu, Bo Du, Wen Yang, Marcello Pelillo, and Liangpei Zhang. Se- mantic change detection with asymmetric siamese net- works. arXiv preprint arXiv:2010.05687, 2020. 5

    work page arXiv 2010

  38. [38]

    Lora-composer: Leverag- ing low-rank adaptation for multi-concept customiza- tion in training-free diffusion models

    Yang Yang, Wen Wang, Liang Peng, Chaotian Song, Yao Chen, Hengjia Li, Xiaolong Yang, Qinglin Lu, Deng Cai, Boxi Wu, et al. Lora-composer: Leverag- ing low-rank adaptation for multi-concept customiza- tion in training-free diffusion models. arXiv preprint arXiv:2403.11627, 2024. 2

    work page arXiv 2024

  39. [39]

    IP-Adapter: Text Compatible Image Prompt Adapter for Text-to-Image Diffusion Models

    Huan Ye, Jian Zhang, Shu Liu, Xinzhu Han, and Wen- jun Yang. Ip-adapter: Text compatible image prompt adapter for text-to-image diffusion models. arXiv preprint, arXiv:2308.06721, 2023. 6

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  40. [40]

    Video probabilistic diffusion models in pro- jected latent space

    Sihyun Yu, Kihyuk Sohn, Subin Kim, and Jinwoo Shin. Video probabilistic diffusion models in pro- jected latent space. In Proceedings of the IEEE/CVF conference on computer vision and pattern recogni- tion, pages 18456–18466, 2023. 2

    work page 2023

  41. [41]

    Metaearth: A generative founda- tion model for global-scale remote sensing image gen- eration

    Zhiping Yu, Chenyang Liu, Liqin Liu, Zhenwei Shi, and Zhengxia Zou. Metaearth: A generative founda- tion model for global-scale remote sensing image gen- eration. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. 1

    work page 2024

  42. [42]

    Changediff: A multi- temporal change detection data generator with flexible text prompts via diffusion model, 2024

    Qi Zang, Jiayi Yang, Shuang Wang, Dong Zhao, Wen- jun Yi, and Zhun Zhong. Changediff: A multi- temporal change detection data generator with flexible text prompts via diffusion model, 2024. 1, 3

    work page 2024

  43. [43]

    Un- balanced feature transport for exemplar-based image translation

    Fangneng Zhan, Yingchen Yu, Kaiwen Cui, Gongjie Zhang, Shijian Lu, Jianxiong Pan, Changgong Zhang, Feiying Ma, Xuansong Xie, and Chunyan Miao. Un- balanced feature transport for exemplar-based image translation. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 15028–15038, 2021. 6

    work page 2021

  44. [44]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision , pages 3836– 3847, 2023. 5, 6

    work page 2023

  45. [45]

    Inversion-based style transfer with diffusion models

    Yuxin Zhang, Nisha Huang, Fan Tang, Haibin Huang, Chongyang Ma, Weiming Dong, and Changsheng Xu. Inversion-based style transfer with diffusion models. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 10146– 10156, 2023. 2

    work page 2023

  46. [46]

    Change is everywhere: Single-temporal supervised object change detection in remote sensing imagery

    Zhuo Zheng, Ailong Ma, Liangpei Zhang, and Yan- fei Zhong. Change is everywhere: Single-temporal supervised object change detection in remote sensing imagery. In Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision (ICCV), pages 15193–15202. IEEE, 2021. 7

    work page 2021

  47. [47]

    Scalable multi-temporal re- mote sensing change data generation via simulating stochastic change process

    Zhuo Zheng, Shiqi Tian, Ailong Ma, Liangpei Zhang, and Yanfei Zhong. Scalable multi-temporal re- mote sensing change data generation via simulating stochastic change process. In Proceedings of the IEEE/CVF International Conference on Computer Vi- sion, pages 21818–21827, 2023. 3

    work page 2023

  48. [48]

    Changen2: Multi-temporal remote sensing generative change foundation model

    Zhuo Zheng, Stefano Ermon, Dongjun Kim, Liangpei Zhang, and Yanfei Zhong. Changen2: Multi-temporal remote sensing generative change foundation model. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. 1, 3

    work page 2024

  49. [49]

    Denoising diffusion bridge models.arXiv preprint arXiv:2309.16948,

    Linqi Zhou, Aaron Lou, Samar Khanna, and Stefano Ermon. Denoising diffusion bridge models. arXiv preprint arXiv:2309.16948, 2023. 2 11

    work page arXiv 2023