pith. sign in

arxiv: 2606.20110 · v1 · pith:EOWIDIE3new · submitted 2026-06-18 · 💻 cs.CV

FrozenDrive: Zero-Shot Text-Guided Driving Scene Generation and Data Augmentation with Parameter-Free Frozen Diffusion Model

Yuhwan Jeong , Hyeonseong Kim , Daehyun We , Seonkyu Song , Jinnyeong Yang , Hyun-Kurl Jang , Youngho Yoon , Kuk-Jin Yoon This is my paper

Pith reviewed 2026-06-26 18:32 UTC · model grok-4.3

classification 💻 cs.CV
keywords diffusion modelsdriving scene generationzero-shot synthesisdata augmentationmulti-view consistencyautonomous drivingtext-guided generationfrozen models
0
0 comments X

The pith

A frozen diffusion model with knowledge-preserving spatio-temporal attention generates coherent multi-view driving scenes from text without fine-tuning or parameter updates.

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

The paper introduces FrozenDrive as a controllable framework that keeps a pretrained diffusion model entirely frozen while using a new attention mechanism to enforce consistency across camera views and time steps. Text prompts combined with driving-stack signals guide the generation, and an object-focused constraint boosts fidelity on rare classes. This produces globally coherent scenes especially in adverse conditions like night or rain. When the synthetic data augments nuScenes training sets, downstream autonomous driving models show clear gains in those same challenging regimes.

Core claim

FrozenDrive conditions on rich driving-stack signals and text prompts, and introduces knowledge-preserving spatio-temporal attention to impose cross-view alignment and temporal coherence in a single pass within a parameter-free frozen diffusion backbone. An additional object-focused constraint improves per-object fidelity for rare categories. Without any weather- or scene-specific fine-tuning, the model synthesizes globally coherent multi-view driving scenes from text, particularly under adverse and rare conditions, and surpasses prior baselines. On nuScenes, FrozenDrive augmented data significantly improves AD models performance, especially at night and in rain.

What carries the argument

Knowledge-preserving spatio-temporal attention mechanism that imposes cross-view alignment and temporal coherence while preserving the pre-trained diffusion model's knowledge and text alignment without any parameter updates or added layers.

If this is right

  • The approach enables synthesis of driving scenes under adverse weather and rare configurations without any fine-tuning.
  • Data generated by the model augments training sets and raises autonomous driving model accuracy, especially at night and in rain.
  • Text alignment and per-object fidelity for infrequent classes remain higher than in fine-tuned alternatives.
  • A single forward pass produces globally consistent multi-view output that exceeds prior methods on the same benchmarks.
  • The object-focused constraint directly lifts fidelity on rare object categories in the generated scenes.

Where Pith is reading between the lines

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

  • The same frozen consistency mechanism could be tested on other multi-camera or video domains where retraining diffusion models is costly.
  • Targeted text prompts might let practitioners generate specific edge-case scenarios on demand rather than relying on real-world collection.
  • If the attention module generalizes, similar lightweight additions could extend to other pre-trained generative backbones without domain-specific retraining.
  • Downstream gains on nuScenes suggest the generated data fills distribution gaps that real datasets under-sample.

Load-bearing premise

The added spatio-temporal attention can enforce multi-view and temporal consistency while leaving the original model's knowledge and text alignment completely intact.

What would settle it

A direct comparison showing that scenes generated by the frozen model with the new attention still lack cross-view or temporal coherence, or that AD models trained on the augmented data show no improvement over baselines in night or rain conditions on nuScenes.

Figures

Figures reproduced from arXiv: 2606.20110 by Daehyun We, Hyeonseong Kim, Hyun-Kurl Jang, Jinnyeong Yang, Kuk-Jin Yoon, Seonkyu Song, Youngho Yoon, Yuhwan Jeong.

Figure 1
Figure 1. Figure 1: (a) Fine-tuning or adding learnable layers lose pre-trained priors and fails to reflect unseen text prompts while training the model. (b) By completely freezing the pretrained model without introducing any additional parameters, our method preserves its rich generative priors, enabling successful zero-shot text-guided generation. As a result, real-world corpora are biased toward frequent, benign conditions… view at source ↗
Figure 2
Figure 2. Figure 2: Overall framework. (a) FrozenDrive uses five conditions: (i) an HD map, (ii) per-view camera indicator, (iii) depth map, (iv) relative pose, and (v) text descrip￾tion. The first four are multi-view, pixel- and spatially aligned and injected into the diffusion model via ControlNet, while the last enables zero-shot text guidance. After the encoder, we replace original self-attention into our knowledge-preser… view at source ↗
Figure 3
Figure 3. Figure 3: Details of the class-wise weighting mask. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative comparison of generated samples [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Visualization of sequential generation. FrozenDrive produces high-fidelity, temporally coherent images across scenes. The rare construction vehicle (red boxes) is faithfully preserved in shape, color, and position across viewpoints and consecutive frames, demonstrating strong object-level consistency. 4 Experiments 4.1 Experimental settings We train and evaluate our proposed method on the nuScenes [6] data… view at source ↗
Figure 6
Figure 6. Figure 6: Examples of text-driven weather-augmented samples. [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Zero-shot text-guided generation results. [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
read the original abstract

Synthetic data for autonomous driving is surging, powered by diffusion models that promise scalable scene generation. Yet key obstacles remain, as enforcing multi-view and temporal consistency often relies on backbone fine-tuning or added layers, which erodes pre-trained knowledge and weakens text alignment. Models also stay close to the training distribution, struggling under adverse weather and unseen configurations, and fidelity favors frequent over rare classes. We address these gaps with FrozenDrive, a controllable generative framework that preserves a pretrained diffusion models knowledge while achieving strong consistency. FrozenDrive conditions on rich driving-stack signals and text prompts, and introduces knowledge-preserving spatio-temporal attention to impose cross-view alignment and temporal coherence in a single pass within a parameter-free frozen diffusion backbone. An additional object-focused constraint improves per-object fidelity for rare categories. Without any weather- or scene-specific fine-tuning, our model synthesizes globally coherent multi-view driving scenes from text, particularly under adverse and rare conditions, and surpasses prior baselines. On nuScenes, FrozenDrive augmented data significantly improves AD models performance, especially at night and in rain, demonstrating stronger robustness when trained with our scenario-targeted data.

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 presents FrozenDrive, a framework for zero-shot text-guided generation of multi-view driving scenes using a frozen pre-trained diffusion model. It introduces a knowledge-preserving spatio-temporal attention mechanism to achieve cross-view and temporal consistency without any parameter updates or additional layers, along with an object-focused constraint to improve fidelity for rare categories. The paper claims that this approach generates globally coherent scenes particularly under adverse and rare conditions, surpasses prior baselines, and when used to augment nuScenes data, significantly improves the performance of autonomous driving models, especially at night and in rain.

Significance. If the central claims hold, the work would be significant for the field of synthetic data generation for autonomous driving. By enabling controllable, consistent multi-view scene synthesis without fine-tuning a diffusion backbone, it could facilitate the creation of targeted data for rare and adverse scenarios while preserving the generative capabilities and text alignment of pre-trained models. This has potential to enhance the robustness of AD systems through data augmentation, as demonstrated on nuScenes.

major comments (2)
  1. [Abstract] Abstract: The abstract asserts that the model 'surpasses prior baselines' and that 'FrozenDrive augmented data significantly improves AD models performance' but provides no quantitative metrics, baselines, error bars, or ablation details to support these claims. This makes the performance assertions difficult to evaluate and is load-bearing for the central superiority claim.
  2. [Method] Method (knowledge-preserving spatio-temporal attention): The mechanism is described as imposing cross-view alignment and temporal coherence in a single pass on a frozen backbone without parameter updates or added layers. Standard diffusion attention operates within single-view latents; no equation, derivation, or analysis is supplied showing how pre-existing attention maps already contain the necessary geometric and temporal correspondences for arbitrary text prompts (including adverse conditions). This assumption is load-bearing for the zero-shot consistency and downstream AD improvement claims.
minor comments (1)
  1. [Abstract] Abstract: The term 'rich driving-stack signals' is used without definition or example, which reduces clarity for readers not already familiar with the conditioning inputs.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We appreciate the referee's thorough review and constructive feedback on our manuscript. Below, we provide point-by-point responses to the major comments.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The abstract asserts that the model 'surpasses prior baselines' and that 'FrozenDrive augmented data significantly improves AD models performance' but provides no quantitative metrics, baselines, error bars, or ablation details to support these claims. This makes the performance assertions difficult to evaluate and is load-bearing for the central superiority claim.

    Authors: The full paper provides detailed quantitative evaluations, including comparisons to baselines with specific metrics, error bars from multiple runs, and ablation studies in Sections 4.2 and 4.3. However, we agree that the abstract would benefit from including key quantitative results to make the claims more concrete. We will revise the abstract to incorporate representative metrics, such as the improvement percentages on nuScenes under adverse conditions. revision: yes

  2. Referee: [Method] Method (knowledge-preserving spatio-temporal attention): The mechanism is described as imposing cross-view alignment and temporal coherence in a single pass on a frozen backbone without parameter updates or added layers. Standard diffusion attention operates within single-view latents; no equation, derivation, or analysis is supplied showing how pre-existing attention maps already contain the necessary geometric and temporal correspondences for arbitrary text prompts (including adverse conditions). This assumption is load-bearing for the zero-shot consistency and downstream AD improvement claims.

    Authors: The manuscript describes the knowledge-preserving spatio-temporal attention as a modification to the attention mechanism that incorporates multi-view and temporal information by attending across the concatenated latents from different views and frames, leveraging the pre-trained model's ability to handle such correspondences through the conditioning signals. We acknowledge that a more formal derivation and analysis of why the pre-existing attention maps suffice for consistency in zero-shot settings would strengthen the presentation. We will add equations detailing the attention computation and an analysis section or appendix discussing the geometric correspondences. revision: yes

Circularity Check

0 steps flagged

No significant circularity; claims rest on described novel mechanism without reduction to inputs or self-citations

full rationale

The provided manuscript text introduces FrozenDrive via a parameter-free knowledge-preserving spatio-temporal attention applied to a frozen diffusion backbone, with no equations, derivations, or fitted parameters shown that reduce any prediction or consistency claim to a self-referential fit. No self-citations are invoked as load-bearing for uniqueness or ansatzes, and the central technique is presented as an independent conditioning approach rather than a renaming or construction from the target outputs. The derivation chain is therefore self-contained against external benchmarks, consistent with the absence of any quoted reduction steps.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that pre-trained diffusion knowledge can be preserved and extended via attention conditioning alone.

axioms (1)
  • domain assumption Pre-trained diffusion models retain useful knowledge when kept frozen and conditioned with driving-stack signals and text prompts.
    Invoked to justify zero-shot performance without fine-tuning.

pith-pipeline@v0.9.1-grok · 5762 in / 1111 out tokens · 17980 ms · 2026-06-26T18:32:28.505490+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

124 extracted references · 2 canonical work pages

  1. [1]

    International Journal of Computer Vision126(9), 961–972 (2018) 4

    Abu Alhaija, H., Mustikovela, S.K., Mescheder, L., Geiger, A., Rother, C.: Aug- mented reality meets computer vision: Efficient data generation for urban driving scenes. International Journal of Computer Vision126(9), 961–972 (2018) 4

    2018

  2. [2]

    arXiv preprint arXiv:2008.03156 (2020) 13

    Aghajanyan, A., Shrivastava, A., Gupta, A., Goyal, N., Zettlemoyer, L., Gupta, S.: Better fine-tuning by reducing representational collapse. arXiv preprint arXiv:2008.03156 (2020) 13

    arXiv 2008

  3. [3]

    arXiv preprint arXiv:2408.06071 (2024) 4

    Assion, F., Gressner, F., Augustine, N., Klemenc, J., Hammam, A., Krattinger, A., Trittenbach, H., Philippsen, A., Riemer, S.: A-bdd: Leveraging data aug- mentations for safe autonomous driving in adverse weather and lighting. arXiv preprint arXiv:2408.06071 (2024) 4

    arXiv 2024

  4. [4]

    arXiv preprint arXiv:2311.15127 (2023) 2, 4

    Blattmann, A., Dockhorn, T., Kulal, S., Mendelevitch, D., Kilian, M., Lorenz, D., Levi, Y., English, Z., Voleti, V., Letts, A., et al.: Stable video diffusion: Scaling latent video diffusion models to large datasets. arXiv preprint arXiv:2311.15127 (2023) 2, 4

    Pith/arXiv arXiv 2023

  5. [5]

    In: CVPR (2023) 4

    Brooks, T., Holynski, A., Efros, A.A.: Instructpix2pix: Learning to follow image editing instructions. In: CVPR (2023) 4

    2023

  6. [6]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Caesar, H., Bankiti, V., Lang, A.H., Vora, S., Liong, V.E., Xu, Q., Krishnan, A., Pan,Y.,Baldan,G.,Beijbom,O.:nuscenes:Amultimodaldatasetforautonomous driving. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 11621–11631 (2020) 2, 9, 10, 16, 17, 18

    2020

  7. [7]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Caron, M., Touvron, H., Misra, I., Jégou, H., Mairal, J., Bojanowski, P., Joulin, A.: Emerging properties in self-supervised vision transformers. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 9650–9660 (2021) 14

    2021

  8. [8]

    Author et al

    Chen, Y., Rong, F., Duggal, S., Wang, S., Yan, X., Manivasagam, S., Xue, S., Yumer, E., Urtasun, R.: Geosim: Realistic video simulation via geometry-aware 26 F. Author et al. nuScenesDriveArena MD-V2 Ours Ours MD-V2 DriveArena nuScenes Ours MD-V2 DriveArena nuScenes Fig.H:Qualitative comparison of generated samples. MD-V2 denotes MagicDrive- V2 [13]. Abbr...

    2021

  9. [9]

    In: ICASSP 2025- 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

    Cheng, H., Xu, J., Peng, L., Yang, Z., He, X., Wu, B.: Object-level data augmen- tation for visual 3d object detection in autonomous driving. In: ICASSP 2025- 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). pp. 1–5. IEEE (2025) 4

    2025

  10. [10]

    arXiv preprint arXiv:2504.18448 (2025) 4

    Dong, H., Wang, X., Lin, D., Wu, Y., Chen, Q., Liu, R., Yang, K., Li, P., Guo, Q.: Noisecontroller: Towards consistent multi-view video generation via noise de- composition and collaboration. arXiv preprint arXiv:2504.18448 (2025) 4

    arXiv 2025

  11. [11]

    arXiv preprint arXiv:2401.03771 (2024) 4

    Feldmann, C., Siegenheim, N., Hars, N., Rabuzin, L., Ertugrul, M., Wolfart, L., Pollefeys, M., Bauer, Z., Oswald, M.R.: Nerfmentation: Nerf-based augmentation for monocular depth estimation. arXiv preprint arXiv:2401.03771 (2024) 4

    arXiv 2024

  12. [12]

    arXiv preprint arXiv:2405.14475 (2024) 2, 4

    Gao, R., Chen, K., Li, Z., Hong, L., Li, Z., Xu, Q.: Magicdrive3d: Control- lable 3d generation for any-view rendering in street scenes. arXiv preprint arXiv:2405.14475 (2024) 2, 4

    arXiv 2024

  13. [13]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Gao, R., Chen, K., Xiao, B., Hong, L., Li, Z., Xu, Q.: Magicdrive-v2: High- resolution long video generation for autonomous driving with adaptive control. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 28135–28144 (2025) 2, 4, 9, 10, 11, 13, 14, 17, 18, 26

    2025

  14. [14]

    In: The Twelfth Inter- national Conference on Learning Representations (2024),https://openreview

    Gao, R., Chen, K., Xie, E., HONG, L., Li, Z., Yeung, D.Y., Xu, Q.: Magicdrive: Street view generation with diverse 3d geometry control. In: The Twelfth Inter- national Conference on Learning Representations (2024),https://openreview. net/forum?id=sBQwvucduK2, 3, 4, 5, 9, 10, 23

    2024

  15. [15]

    Advances in Neural Information Processing Systems37, 91560–91596 (2024)

    Gao, S., Yang, J., Chen, L., Chitta, K., Qiu, Y., Geiger, A., Zhang, J., Li, H.: Vista: A generalizable driving world model with high fidelity and versatile con- trollability. Advances in Neural Information Processing Systems37, 91560–91596 (2024)

    2024

  16. [16]

    arXiv preprint arXiv:2506.11526 (2025)

    Gao, Y., Piccinini, M., Zhang, Y., Wang, D., Moller, K., Brusnicki, R., Zarrouki, B., Gambi, A., Totz, J.F., Storms, K., et al.: Foundation models in autonomous driving: A survey on scenario generation and scenario analysis. arXiv preprint arXiv:2506.11526 (2025)

    arXiv 2025

  17. [17]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Gatys, L.A., Ecker, A.S., Bethge, M.: Image style transfer using convolutional neural networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 2414–2423 (2016) 4

    2016

  18. [18]

    arXiv preprint arXiv:2503.18108 (2025)

    Ge, J., Liu, Z., Fan, L., Jiang, Y., Su, J., Li, Y., Zhang, Z., Chen, S.: Unraveling the effects of synthetic data on end-to-end autonomous driving. arXiv preprint arXiv:2503.18108 (2025)

    arXiv 2025

  19. [19]

    IEEE Robotics and Automation Letters6(3), 4947– 4954 (2021) 22

    Gehrig, M., Aarents, W., Gehrig, D., Scaramuzza, D.: Dsec: A stereo event camera dataset for driving scenarios. IEEE Robotics and Automation Letters6(3), 4947– 4954 (2021) 22

    2021

  20. [20]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Gu, J., Hu, C., Zhang, T., Chen, X., Wang, Y., Wang, Y., Zhao, H.: Vip3d: End-to-end visual trajectory prediction via 3d agent queries. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 5496–5506 (2023) 1

    2023

  21. [21]

    arXiv preprint arXiv:2503.15208 (2025) 2, 4, 9, 10, 23

    Guo, J., Ding, Y., Chen, X., Chen, S., Li, B., Zou, Y., Lyu, X., Tan, F., Qi, X., Li, Z., et al.: Dist-4d: Disentangled spatiotemporal diffusion with metric depth for 4d driving scene generation. arXiv preprint arXiv:2503.15208 (2025) 2, 4, 9, 10, 23

    arXiv 2025

  22. [22]

    Author et al

    Guo, X., Wu, Z., Xiong, K., Xu, Z., Zhou, L., Xu, G., Xu, S., Sun, H., WANG, B., Chen, G., Ye, H., Liu, W., Wang, X.: Genesis: Multimodal driving scene 28 F. Author et al. generation with spatio-temporal and cross-modal consistency. In: The Thirty- ninth Annual Conference on Neural Information Processing Systems (2025), https://openreview.net/forum?id=Q7YnqREWLq4

    2025

  23. [23]

    In: Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume

    He, T., Liu, J., Cho, K., Ott, M., Liu, B., Glass, J., Peng, F.: Analyzing the forget- ting problem in pretrain-finetuning of open-domain dialogue response models. In: Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume. pp. 1121–1133 (2021) 5, 13

    2021

  24. [24]

    arXiv preprint arXiv:2211.13221 (2022) 4

    He, Y., Yang, T., Zhang, Y., Shan, Y., Chen, Q.: Latent video diffusion models for high-fidelity long video generation. arXiv preprint arXiv:2211.13221 (2022) 4

    Pith/arXiv arXiv 2022

  25. [25]

    In: The Eleventh International Conference on Learning Representations (2023),https: //openreview.net/forum?id=_CDixzkzeyb4

    Hertz, A., Mokady, R., Tenenbaum, J., Aberman, K., Pritch, Y., Cohen- Or, D.: Prompt-to-prompt image editing with cross-attention control. In: The Eleventh International Conference on Learning Representations (2023),https: //openreview.net/forum?id=_CDixzkzeyb4

    2023

  26. [26]

    Advances in neural information processing systems33, 6840–6851 (2020) 4, 8

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. Advances in neural information processing systems33, 6840–6851 (2020) 4, 8

    2020

  27. [27]

    arXiv preprint arXiv:2207.12598 (2022) 4

    Ho, J., Salimans, T.: Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598 (2022) 4

    Pith/arXiv arXiv 2022

  28. [28]

    arXiv preprint arXiv:2309.17080 (2023) 4

    Hu, A., Russell, L., Yeo, H., Murez, Z., Fedoseev, G., Kendall, A., Shotton, J., Corrado, G.: Gaia-1: A generative world model for autonomous driving. arXiv preprint arXiv:2309.17080 (2023) 4

    Pith/arXiv arXiv 2023

  29. [29]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Hu, Y., Yang, J., Chen, L., Li, K., Sima, C., Zhu, X., Chai, S., Du, S., Lin, T., Wang, W., et al.: Planning-oriented autonomous driving. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 17853– 17862 (2023) 1, 3, 9, 10, 18

    2023

  30. [30]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Huang, Z., He, Y., Yu, J., Zhang, F., Si, C., Jiang, Y., Zhang, Y., Wu, T., Jin, Q., Chanpaisit, N., et al.: Vbench: Comprehensive benchmark suite for video generative models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 21807–21818 (2024) 14

    2024

  31. [31]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Islam, K., Zaheer, M.Z., Mahmood, A., Nandakumar, K.: Diffusemix: Label- preserving data augmentation with diffusion models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 27621– 27630 (2024) 2, 4

    2024

  32. [32]

    Advances in Neural Information Processing Systems37, 74211–74232 (2024)

    Jang, H.K., Kim, J., Kweon, H., Yoon, K.J.: Talos: Enhancing semantic scene completion via test-time adaptation on the line of sight. Advances in Neural Information Processing Systems37, 74211–74232 (2024)

    2024

  33. [33]

    In: CVPR (2020)

    Jaritz, M., Vu, T.H., de Charette, R., Wirbel, E., Pérez, P.: xMUDA: Cross-modal unsupervised domain adaptation for 3D semantic segmentation. In: CVPR (2020)

    2020

  34. [34]

    In: European Conference on Computer Vision

    Jeong, Y., Cho, H., Yoon, K.J.: Towards robust event-based networks for night- time via unpaired day-to-night event translation. In: European Conference on Computer Vision. pp. 286–306. Springer (2024)

    2024

  35. [35]

    arXiv preprint arXiv:2503.22231 (2025) 4

    Ji, Y., Zhu, Z., Zhu, Z., Xiong, K., Lu, M., Li, Z., Zhou, L., Sun, H., Wang, B., Lu, T.: Cogen: 3d consistent video generation via adaptive conditioning for autonomous driving. arXiv preprint arXiv:2503.22231 (2025) 4

    arXiv 2025

  36. [36]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Jiang, B., Chen, S., Xu, Q., Liao, B., Chen, J., Zhou, H., Zhang, Q., Liu, W., Huang, C., Wang, X.: Vad: Vectorized scene representation for efficient au- tonomous driving. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 8340–8350 (2023) 1

    2023

  37. [37]

    arXiv preprint arXiv:2504.19614 (2025) 4 Abbreviated paper title 29

    Jiang, J., Hong, G., Zhang, M., Hu, H., Zhan, K., Shao, R., Nie, L.: Dive: Efficient multi-view driving scenes generation based on video diffusion transformer. arXiv preprint arXiv:2504.19614 (2025) 4 Abbreviated paper title 29

    arXiv 2025

  38. [38]

    Advances in neural information processing systems35, 23593–23606 (2022)

    Kawar, B., Elad, M., Ermon, S., Song, J.: Denoising diffusion restoration models. Advances in neural information processing systems35, 23593–23606 (2022)

    2022

  39. [39]

    ACM Trans

    Kerbl, B., Kopanas, G., Leimkühler, T., Drettakis, G.: 3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph.42(4), 139–1 (2023) 4

    2023

  40. [40]

    In: NeurIPS 2023 Workshop on Distribution Shifts: New Frontiers with Foundation Models (2024),https: //openreview.net/forum?id=wkQy8mLIb95

    Kotha, S., Springer, J., Raghunathan, A.: Understanding catastrophic forget- ting in language models via implicit inference. In: NeurIPS 2023 Workshop on Distribution Shifts: New Frontiers with Foundation Models (2024),https: //openreview.net/forum?id=wkQy8mLIb95

    2023

  41. [41]

    In: Proceedings of the IEEE/CVF confer- ence on computer vision and pattern recognition

    Kumari, N., Zhang, B., Zhang, R., Shechtman, E., Zhu, J.Y.: Multi-concept cus- tomization of text-to-image diffusion. In: Proceedings of the IEEE/CVF confer- ence on computer vision and pattern recognition. pp. 1931–1941 (2023) 5, 13

    1931

  42. [42]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Li, B., Guo, J., Liu, H., Zou, Y., Ding, Y., Chen, X., Zhu, H., Tan, F., Zhang, C., Wang, T., et al.: Uniscene: Unified occupancy-centric driving scene generation. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 11971–11981 (2025) 10, 19

    2025

  43. [43]

    In: IEEE International Conference on Robotics and Automation (ICRA)

    Li, H., Yang, Z., Qian, Z., Zhao, G., Huang, Y., Yu, J., Zhou, H., Liu, L.: Duald- iff: Dual-branch diffusion model for autonomous driving with semantic fusion. In: IEEE International Conference on Robotics and Automation (ICRA). IEEE (2025),https://arxiv.org/abs/2505.018574

    arXiv 2025

  44. [44]

    In: Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition

    Li, J., Li, B., Tu, Z., Liu, X., Guo, Q., Juefei-Xu, F., Xu, R., Yu, H.: Light the night: A multi-condition diffusion framework for unpaired low-light enhancement in autonomous driving. In: Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition. pp. 15205–15215 (2024) 2, 4

    2024

  45. [45]

    In: 2022 international conference on robotics and automation (ICRA)

    Li, N., Song, F., Zhang, Y., Liang, P., Cheng, E.: Traffic context aware data aug- mentation for rare object detection in autonomous driving. In: 2022 international conference on robotics and automation (ICRA). pp. 4548–4554. IEEE (2022) 4

    2022

  46. [46]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Li, R., Cheong, L.F., Tan, R.T.: Heavy rain image restoration: Integrating physics model and conditional adversarial learning. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 1633–1642 (2019) 11

    2019

  47. [47]

    In: European Conference on Computer Vision

    Li, X., Zhang, Y., Ye, X.: Drivingdiffusion: layout-guided multi-view driving sce- narios video generation with latent diffusion model. In: European Conference on Computer Vision. pp. 469–485. Springer (2024) 4

    2024

  48. [48]

    arXiv preprint arXiv:2103.05422 (2021) 4

    Li, X., Kou, K., Zhao, B.: Weather gan: Multi-domain weather translation using generative adversarial networks. arXiv preprint arXiv:2103.05422 (2021) 4

    arXiv 2021

  49. [49]

    In: Proceedings of the ieee/cvf in- ternational conference on computer vision

    Li, Y., Lin, Z.H., Forsyth, D., Huang, J.B., Wang, S.: Climatenerf: Extreme weather synthesis in neural radiance field. In: Proceedings of the ieee/cvf in- ternational conference on computer vision. pp. 3227–3238 (2023) 4

    2023

  50. [50]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Li, Z., Li, L., Zhu, J.: Read: Large-scale neural scene rendering for autonomous driving. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 37, pp. 1522–1529 (2023) 4

    2023

  51. [51]

    arXiv preprint arXiv:2203.04600 (2022) 5

    Li, Z., Ren, K., Jiang, X., Li, B., Zhang, H., Li, D.: Domain generalization using pretrained models without fine-tuning. arXiv preprint arXiv:2203.04600 (2022) 5

    arXiv 2022

  52. [52]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Liang, Y., Yan, Z., Chen, L., Zhou, J., Yan, L., Zhong, S., Zou, X.: Driveeditor: A unified 3d information-guided framework for controllable object editing in driving scenes. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 39, pp. 5164–5172 (2025) 4

    2025

  53. [53]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Liao, Y.C., Chen, J.J., Huang, C.P., Lin, C.S., Wu, M.L., Wang, Y.C.F.: Con- tinual personalization for diffusion models. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 15511–15520 (2025) 5, 13 30 F. Author et al

    2025

  54. [54]

    In: Proceedings of the 32nd ACM International Conference on Multimedia

    Liu, B., Wang, K., Liu, Y., Bao, J., Han, T., Yu, J.: Mvpbev: Multi-view perspec- tive image generation from bev with test-time controllability and generalizability. In: Proceedings of the 32nd ACM International Conference on Multimedia. pp. 8393–8401 (2024) 4

    2024

  55. [55]

    Loshchilov,I.,Hutter,F.:Decoupledweightdecayregularization.In:International Conference on Learning Representations (2019) 17

    2019

  56. [56]

    arXiv preprint arXiv:2412.03520 (2024) 4

    Lu, H., Wu, X., Wang, S., Qin, X., Zhang, X., Han, J., Zuo, W., Tao, J.: Seeing beyond views: Multi-view driving scene video generation with holistic attention. arXiv preprint arXiv:2412.03520 (2024) 4

    arXiv 2024

  57. [57]

    In: European Conference on Computer Vision

    Lu, J., Huang, Z., Yang, Z., Zhang, J., Zhang, L.: Wovogen: World volume-aware diffusion for controllable multi-camera driving scene generation. In: European Conference on Computer Vision. pp. 329–345. Springer (2024) 4

    2024

  58. [58]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Lu,Y.,Ren,X.,Yang,J., Shen,T., Wu,Z.,Gao,J., Wang, Y.,Chen, S.,Chen, M., Fidler,S.,etal.:Infinicube:Unboundedandcontrollabledynamic3ddrivingscene generation with world-guided video models. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 27272–27283 (2025) 4

    2025

  59. [59]

    arXiv preprint arXiv:2409.04003 (2024) 4

    Mei, J., Hu, T., Yang, X., Wen, L., Yang, Y., Wei, T., Ma, Y., Dou, M., Shi, B., Liu, Y.: Dreamforge: Motion-aware autoregressive video generation for multi-view driving scenes. arXiv preprint arXiv:2409.04003 (2024) 4

    arXiv 2024

  60. [60]

    In: International Conference on Learning Representations (2022) 4

    Meng, C., He, Y., Song, Y., Song, J., Wu, J., Zhu, J.Y., Ermon, S.: SDEdit: Guided image synthesis and editing with stochastic differential equations. In: International Conference on Learning Representations (2022) 4

    2022

  61. [61]

    Com- munications of the ACM65(1), 99–106 (2021) 4

    Mildenhall, B., Srinivasan, P.P., Tancik, M., Barron, J.T., Ramamoorthi, R., Ng, R.: Nerf: Representing scenes as neural radiance fields for view synthesis. Com- munications of the ACM65(1), 99–106 (2021) 4

    2021

  62. [62]

    Ming, Y., Li, Y.: How does fine-tuning impact out-of-distribution detection for vision-language models? International Journal of Computer Vision132(2), 596– 609 (2024) 5

    2024

  63. [63]

    arXiv preprint arXiv:2308.13320 (2023) 13

    Mukhoti, J., Gal, Y., Torr, P.H., Dokania, P.K.: Fine-tuning can cripple your foundation model; preserving features may be the solution. arXiv preprint arXiv:2308.13320 (2023) 13

    arXiv 2023

  64. [64]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Ni, C., Zhao, G., Wang, X., Zhu, Z., Qin, W., Huang, G., Liu, C., Chen, Y., Wang, Y., Zhang, X., et al.: Recondreamer: Crafting world models for driving scene reconstruction via online restoration. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 1559–1569 (2025) 4

    2025

  65. [65]

    arXiv preprint arXiv:2409.08248 (2024) 5, 13

    Park, N., Kim, K., Shim, H.: Textboost: Towards one-shot personalization of text- to-image models via fine-tuning text encoder. arXiv preprint arXiv:2409.08248 (2024) 5, 13

    Pith/arXiv arXiv 2024

  66. [66]

    The International Journal of Robotics Research40(4-5), 681–690 (2021) 22

    Pitropov, M., Garcia, D.E., Rebello, J., Smart, M., Wang, C., Czarnecki, K., Waslander, S.: Canadian adverse driving conditions dataset. The International Journal of Robotics Research40(4-5), 681–690 (2021) 22

    2021

  67. [67]

    Lambeta, P.-W

    Qian, C., Guo, Y., Mo, Y., Li, W.: Weatherdg: Llm-assisted procedural weather generation for domain-generalized semantic segmentation. IEEE Robotics and Automation Letters10(6), 5919–5926 (2025).https://doi.org/10.1109/LRA. 2025.35598214

    work page doi:10.1109/lra 2025

  68. [68]

    In: International conference on machine learning

    Radford,A.,Kim,J.W.,Hallacy,C.,Ramesh,A.,Goh,G.,Agarwal,S.,Sastry,G., Askell, A., Mishkin, P., Clark, J., et al.: Learning transferable visual models from natural language supervision. In: International conference on machine learning. pp. 8748–8763. PmLR (2021) 14, 19 Abbreviated paper title 31

    2021

  69. [69]

    In: 2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

    Ram, S., Neiman, T., Feng, Q., Stuart, A., Tran, S., Chilimbi, T.: Dreamblend: Advancing personalized fine-tuning of text-to-image diffusion models. In: 2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). pp. 3614–3623. IEEE (2025) 13

    2025

  70. [70]

    Ren, X., Lu, Y., Cao, T., Gao, R., Huang, S., Sabour, A., Shen, T., Pfaff, T., Wu, J.Z., Chen, R., et al.: Cosmos-drive-dreams: Scalable synthetic driving data generationwithworldfoundationmodels.arXivpreprintarXiv:2506.09042(2025)

    arXiv 2025

  71. [71]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., Ommer, B.: High-resolution image synthesis with latent diffusion models. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 10684–10695 (2022) 2, 4, 5, 9, 17

    2022

  72. [72]

    In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision

    Rothmeier, T., Huber, W., Knoll, A.C.: Time to shine: Fine-tuning object de- tection models with synthetic adverse weather images. In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. pp. 4447– 4456 (2024) 4

    2024

  73. [73]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Ruiz, N., Li, Y., Jampani, V., Pritch, Y., Rubinstein, M., Aberman, K.: Dream- booth: Fine tuning text-to-image diffusion models for subject-driven generation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 22500–22510 (2023) 13

    2023

  74. [74]

    arXiv preprint arXiv:2503.20523 (2025) 4

    Russell, L., Hu, A., Bertoni, L., Fedoseev, G., Shotton, J., Arani, E., Corrado, G.: Gaia-2: A controllable multi-view generative world model for autonomous driving. arXiv preprint arXiv:2503.20523 (2025) 4

    Pith/arXiv arXiv 2025

  75. [75]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Ryu, K., Hwang, S., Park, J.: Instant domain augmentation for lidar semantic segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 9350–9360 (2023) 4

    2023

  76. [76]

    arXiv preprint arXiv:2310.15110 (2023)

    Shi, R., Chen, H., Zhang, Z., Liu, M., Xu, C., Wei, X., Chen, L., Zeng, C., Su, H.: Zero123++: a single image to consistent multi-view diffusion base model. arXiv preprint arXiv:2310.15110 (2023)

    Pith/arXiv arXiv 2023

  77. [77]

    arXiv:2010.02502 (October 2020),https://arxiv.org/abs/2010.025024

    Song, J., Meng, C., Ermon, S.: Denoising diffusion implicit models. arXiv:2010.02502 (October 2020),https://arxiv.org/abs/2010.025024

    Pith/arXiv arXiv 2010

  78. [78]

    In: International Conference on Learning Representations (2021),https : / / openreview.net/forum?id=PxTIG12RRHS4

    Song, Y., Sohl-Dickstein, J., Kingma, D.P., Kumar, A., Ermon, S., Poole, B.: Score-based generative modeling through stochastic differential equations. In: International Conference on Learning Representations (2021),https : / / openreview.net/forum?id=PxTIG12RRHS4

    2021

  79. [79]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Sun, P., Kretzschmar, H., Dotiwalla, X., Chouard, A., Patnaik, V., Tsui, P., Guo, J., Zhou, Y., Chai, Y., Caine, B., et al.: Scalability in perception for autonomous driving: Waymo open dataset. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 2446–2454 (2020) 2, 22

    2020

  80. [80]

    In: 2025 IEEE International Conference on Robotics and Automation (ICRA)

    Sun, W., Lin, X., Shi, Y., Zhang, C., Wu, H., Zheng, S.: Sparsedrive: End-to-end autonomous driving via sparse scene representation. In: 2025 IEEE International Conference on Robotics and Automation (ICRA). pp. 8795–8801. IEEE (2025) 1, 11, 18

    2025

Showing first 80 references.