Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

DiVE: Efficient Multi-View Driving Scenes Generation Based on Video Diffusion Transformer

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Multi-view driving model lifts 3D detection mAP by 2.9 points

desk verdict A real engineering contribution, but the headline SOTA and augmentation claims rest on uncontrolled cross-paper comparisons, so the numbers are promising, not proven. read the letter →

arxiv 2504.19614 v1 pith:US3ZSQAT submitted 2025-04-28 cs.CV

Junpeng Jiang , Gangyi Hong , Miao Zhang , Hengtong Hu , Kun Zhan , Rui Shao , Liqiang Nie This is my paper · ORCID
classification cs.CV
keywords controllablevideogenerationmulti-viewdiffusiontransformerdrivingscenebird's-eye-viewcontrolmulti-modalconditioningclassifier-freeguidancedistillationprogressiveresolutionsampling
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

DiVE is a video-generation system for autonomous-driving scenes that sets out to establish that a diffusion-transformer generator, conditioned on text, bird's-eye-view road sketches, 3D object boxes, and camera geometry, can produce multi-camera driving videos that are photorealistic, temporally coherent, and consistent across views. The paper argues that such synthetic videos are genuinely useful as training data: adding DiVE-generated clips to real nuScenes data lifts the 3D detection model's mAP from 38.0 to 40.9 and NDS from 49.0 to 52.0. To make high-resolution generation affordable, it introduces two mechanisms—a distillation scheme that replaces per-condition classifier-free guidance selection, and a training-free progressive-resolution sampler—and reports a combined 2.62x speedup with modest quality loss. The stakes are practical: if the central claim holds, controllable generation can lower the cost of collecting and annotating diverse driving video.

What carries the argument

The load-bearing mechanism is a diffusion transformer that interleaves spatial, temporal, cross-attention, and MLP layers, extended by three components. View-inflated attention reshapes multi-view tokens so all cameras attend together without added parameters, enforcing cross-view consistency. A unified cross-attention fuses text tokens, Fourier-encoded 3D object boxes fused with instance captions, and camera-pose embeddings; SketchFormer, a geometry-aware sketch encoder, injects bird's-eye-view road layouts through mirrored zero-initialized fusion cells. Efficiency comes from Multi-Control Auxiliary Branch Distillation (MAD), which trains lightweight condition-specific auxiliary branches to imitate multi-condition classifier-free guidance in one forward pass, and Resolution Progressively Sampling (RPS), which runs early sampling stages at lower resolutions with a resolution-aware timestep shift before refining at full resolution.

What would settle it

Run every compared generator through a single evaluation pipeline—same resolution, frame count, guidance scale, sampling steps, and identical FID/FVD/KPM implementations—then train the same 3D detector on each synthetic set at matched resolution; if DiVE no longer leads on the generation metrics or loses its downstream mAP/NDS gain, the paper's main claims are not supported.

Watch

Extended reading notes

Core claim

The paper's central claim is that DiVE achieves state-of-the-art quality for multi-view driving video generation on nuScenes, with FID 7.14, FVD 68.4, and a cross-view key-point matching score of 73.2, improving over the prior generators it compares against. It further claims that the generated videos are controllable enough to match real-data geometry: BEV road and vehicle segmentation scores and 3D object-detection mAP/NDS approach the real-data reference, and augmenting real training data with DiVE output raises downstream perception performance by a larger margin than prior synthetic generators. The authors state these results as evidence that a diffusion transformer with multi-modal conditioning, cross-view attention, and efficient guidance handling can serve as a practical data simulator for autonomous driving perception.

Load-bearing premise

The comparison against prior methods assumes their published FID, FVD, and KPM scores were measured under the same evaluation protocol; because these metrics depend on resolution, frame count, and the perceptual-metric backbone, a fair re-run could narrow or change the reported margin.

Editorial extensions

If this is right

  • DiVE-generated video becomes a viable augmentation source: real-plus-synthetic training raises the video-based 3D detector's mAP by 2.9 points and NDS by 3.0 points over real data alone.
  • Multi-condition classifier-free guidance can be compressed to a single function evaluation, so combinatorial guidance tuning is replaced by a fixed auxiliary-branch recipe.
  • Parameter-free view-inflated attention gives cross-view consistency at zero added parameters, making it a drop-in change for existing DiT video architectures.
  • Progressive-resolution sampling with a timestep shift yields 2.62x faster high-resolution generation with only small FVD degradation.
  • The generated scenes remain controllable across long horizons, with stable object appearance and layout over 240-frame videos.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: view-inflated attention is a generic token-grouping operation; it should transfer to any multi-view or panoramic video generator, and a direct test is measuring KPM gains when the same reshape is added to a UNet-based generator.
  • Editorial inference: MAD's mixed-control training—randomly nullifying subsets of conditions during distillation—is a portable recipe for multi-condition diffusion models; a layout-to-image model with text and box conditions could validate it outside driving.
  • Editorial inference: the RPS timestep-shift formula is stated for arbitrary resolution ratios and should port to other rectified-flow models as a training-free upsampling sampler; the paper only demonstrates the 240p-to-480p ladder.
  • Editorial inference: if the downstream gains reproduce, the practical endgame is targeted generation of rare driving conditions (night, rain, crowded scenes) rather than collecting them; the paper demonstrates editing ability but does not quantify a corner-case benchmark.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper presents DiVE, a DiT-based framework for generating multi-view driving videos under multimodal control (text, 3D object instances, road sketches, camera parameters). The architecture combines unified cross-attention, a SketchFormer for road guidance, a parameter-free view-inflated attention for cross-view consistency, and a multi-scale training schedule. To reduce inference cost, the authors introduce Multi-Control Auxiliary Branch Distillation (MAD) and Resolution Progressive Sampling (RPS), reporting a combined 2.62x speedup. Experiments on nuScenes report state-of-the-art FID 7.14, FVD 68.4, and KPM 73.2, and show that training StreamPETR with DiVE-generated data plus real data raises mAP from 38.0 to 40.9 and NDS from 49.0 to 52.0, claiming larger gains than Panacea and Glad.

Significance. If validated, DiVE would be a useful step toward controllable, high-resolution synthetic video for autonomous driving perception. The view-inflated attention trick is simple and parameter-free, the MAD/RPS acceleration ideas are practically motivated, and the paper includes extensive ablations and a downstream augmentation study. However, the headline SOTA and augmentation-superiority claims currently rest on cross-paper metric comparisons and on augmentation baselines that differ in resolution and starting performance. The architectural contributions are solid, but the evidence for the strongest claims needs to be re-established under a shared evaluation protocol.

major comments (4)
  1. [Section 4.3, Table 1] The SOTA claim is based on metric values taken from prior papers without a shared evaluation protocol. Table 1 lists resolutions from 192x384 to 480x854 but does not report the number of frames, the FVD feature extractor, or the KPM matching model and thresholds for each method. FVD is sensitive to frame count and backbone, and KPM depends on the matcher and thresholds, so the reported margins (DiVE FVD 68.4 vs Panacea 139.0) cannot be taken at face value. Please either re-run all prior methods under the exact protocol used for DiVE, or provide a detailed protocol table and a sensitivity analysis. Without this, the 'state-of-the-art' claim in Section 4.3 is not fully supported.
  2. [Section 4.4 and Table 1] There is an unexplained discrepancy in DiVE's FVD. Table 1 reports DiVE FVD 68.4 and DiVE+ FVD 93.2, while Figure 7 and Tables 4 and 5 report DiVE FVD 86.13 and DiVE+ FVD 100.18 under a reduced protocol (first 16 frames per scene across four runs). The paper does not state which protocol underlies Table 1 or whether the reduced protocol matches the protocols used for prior methods. Because the headline SOTA margin depends on the 68.4 FVD value, the exact protocol (frames per clip, number of clips, resolution, FVD implementation) must be specified for both the main table and the ablations, and the discrepancy must be explained.
  3. [Table 2 and Appendix B.4] The augmentation comparison is confounded. The DiVE block starts from a higher real-data StreamPETR baseline (38.0 mAP / 49.0 NDS vs 34.6 / 47.0 for Panacea) and uses a different training resolution: Appendix B.4 states that StreamPETR training for DiVE is at 480p, differing from the 256x704 baseline. The +2.9 mAP improvement for DiVE over the +1.7 mAP for Panacea could therefore be due to higher perception training resolution rather than generator quality. To support the claim that DiVE-generated data is superior for augmentation, compare all generators at the same training resolution and from the same real-data baseline, e.g., by upsampling Panacea/Glad data to 480p or downsampling DiVE data to 256x512.
  4. [Figure 7 and Section 4.5] The claim of 'minimal quality degradation' for DiVE+ is not supported by the reported FVD. The FVD increases from 86.13 (DiVE) to 100.18 (DiVE+) in Figure 7, and from 68.4 to 93.2 in Table 1, the latter being a 36% relative increase. Since FVD is the paper's primary temporal-quality metric, 'minimal' should be replaced with a quantitative statement, or task-relevant evidence should be provided showing that the FVD change does not affect downstream perception utility.
minor comments (5)
  1. [Abstract] There are typos: 'coniguration' should be 'configuration', and 'researches' should be 'research'.
  2. [Section 4.2] The sentence 'a guidance scale value is uniformly sampled fron the continous interval' contains two typos: 'fron' and 'continous'.
  3. [Related Work] 'In constrast' should be 'In contrast'.
  4. [Table 2] The '+1.7%' and '+2.9%' entries are absolute percentage-point changes in mAP, not relative percentages; please use 'pp' or write 'mAP +1.7 points' to avoid ambiguity.
  5. [Figure 1 caption] 'CFG denots' should be 'CFG denotes'.

Circularity Check

1 steps flagged · score 2.0 of 10

No significant circularity in the core derivation; one ancillary self-citation in the W-CODA comparison against the authors' earlier DiVE workshop paper is not load-bearing.

  1. other [Section 4.5 (Quantitative Results in W-CODA'24), Figure 8, and reference [22]]
    "DiVE consistently outperforms both UNet-based [8, 11, 34] and DiT-based [10, 22] models across all metrics, with particularly notable advancements in generative quality and fine-grained controllability."

    Reference [22] is the authors' own prior workshop paper, 'DiVE: DiT-based Video Generation with Enhanced Control' (Junpeng Jiang et al., ECCV 2024 Workshop). The W-CODA comparison thus claims superiority over the previous iteration of the same model and research group, so it is not an independent external benchmark. This is a minor self-citation rather than a load-bearing circular step: the main nuScenes SOTA claim in Table 1 and the StreamPETR augmentation claim in Table 2 are supported by external metrics, pretrained perception models, and re-implemented baselines.

full rationale

DiVE's central derivation chain is empirical and externally anchored. The generator is fine-tuned from OpenSora 1.1 with cross-attention conditioning, SketchFormer, and view-inflated attention, and the claimed benefits are evaluated with external measurements (FID, FVD, KPM), pretrained perception models (CVT, BEVFusion, BEVFormer, StreamPETR), and re-implemented baselines (Panacea). No equation or fitted parameter is renamed as a prediction: MAD distills the teacher CFG output exactly as its training target (Algorithm 1, line 9), and RPS's timestep shift in Eq. (3) is a standard resolution-aware noise schedule rather than a fit to the reported FVD. The Table 1 SOTA comparison uses cross-paper numbers and Table 2 mixes StreamPETR training resolutions (Appendix B.4 states 480p for DiVE vs. 256x704 baseline), but these are protocol-comparability concerns, not circular reductions. The only self-referential element is the Figure 8 comparison against the authors' prior workshop DiVE [22]; it is ancillary and does not support the central generation or augmentation claims. Hence the overall circularity is limited to one minor, non-load-bearing self-citation.

Assumptions & free parameters 3 free parameters · 3 assumptions · 3 invented entities

The paper does not derive any constants from first principles; its contribution is empirical and system-level. It relies on pretrained components (OpenSora, VAE, T5, CLIP) and standard generative-modeling assumptions. The main invented entities are architectural modules whose value is shown only through the paper's own ablations, which is normal for this type of work.

free parameters (3)
  • Guidance scale lambda = 2.0 at inference; [1,8] during MAD training
    The CFG guidance scale is a hyperparameter that controls the strength of conditioning; it is chosen and not derived.
  • RPS stage step allocation = 10 steps at 240p, 10 at 360p, 10 at 480p
    The step distribution across resolutions is selected based on an ablation (Table 5) and is not derived from a principle.
  • Number of SketchFormer fusion cells = 13
    The number of fusion cells is a design choice, though it is described as synchronized with the early layers.
assumptions (3)
  • domain assumption The pretrained OpenSora 1.1 DiT provides a strong prior that transfers to driving video generation with only moderate fine-tuning.
    The paper initializes from OpenSora 1.1 weights and assumes the base model's distribution supports driving-domain fine-tuning. This is standard practice in the field.
  • domain assumption The frozen LDM VAE generalizes to 480x854 multi-view driving video latents.
    The paper uses a frozen pretrained VAE to encode multi-view videos; distortion from VAE reconstruction is not measured in the paper.
  • domain assumption Perception metrics (FID, FVD, KPM) computed on generated data with the given feature extractors reflect the usefulness of the data for downstream perception.
    The paper's central quality claims rely on these metrics as proxies for downstream utility. The paper does provide direct downstream evaluation, which partially supports this assumption.
invented entities (3)
  • MAD auxiliary branches with mixed-control guidance training
    purpose: Distill multi-condition classifier-free guidance into a single forward pass with condition-specific auxiliary branches
    This is a methodological contribution whose value is only demonstrated internally on the paper's own benchmark.
  • View-inflated attention
    purpose: Enforce cross-view consistency by reshaping tokens across views before spatial attention
    This is a parameter-free architectural modification; its effectiveness is only shown in the paper's own ablations.
  • SketchFormer
    purpose: Inject BEV road-layout information into the DiT via 13 fusion cells with zero-initialized linear projections
    A new conditioning module inspired by PixArt-delta and ControlNet; it is not independently validated outside this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DiVE: Efficient Multi-View Driving Scenes Generation Based on Video Diffusion Transformer." pith.science (2026). https://pith.science/paper/US3ZSQAT

@misc{pith2026250419614,
  author       = {Pith},
  title        = {Pith review of: DiVE: Efficient Multi-View Driving Scenes Generation Based on Video Diffusion Transformer},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/US3ZSQAT}},
  note         = {Machine review of arXiv:2504.19614}
}
read the original abstract

Collecting multi-view driving scenario videos to enhance the performance of 3D visual perception tasks presents significant challenges and incurs substantial costs, making generative models for realistic data an appealing alternative. Yet, the videos generated by recent works suffer from poor quality and spatiotemporal consistency, undermining their utility in advancing perception tasks under driving scenarios. To address this gap, we propose DiVE, a diffusion transformer-based generative framework meticulously engineered to produce high-fidelity, temporally coherent, and cross-view consistent multi-view videos, aligning seamlessly with bird's-eye view layouts and textual descriptions. DiVE leverages a unified cross-attention and a SketchFormer to exert precise control over multimodal data, while incorporating a view-inflated attention mechanism that adds no extra parameters, thereby guaranteeing consistency across views. Despite these advancements, synthesizing high-resolution videos under multimodal constraints introduces dual challenges: investigating the optimal classifier-free guidance coniguration under intricate multi-condition inputs and mitigating excessive computational latency in high-resolution rendering--both of which remain underexplored in prior researches. To resolve these limitations, we introduce two innovations: Multi-Control Auxiliary Branch Distillation, which streamlines multi-condition CFG selection while circumventing high computational overhead, and Resolution Progressive Sampling, a training-free acceleration strategy that staggers resolution scaling to reduce high latency due to high resolution. These innovations collectively achieve a 2.62x speedup with minimal quality degradation. Evaluated on the nuScenes dataset, DiVE achieves SOTA performance in multi-view video generation, yielding photorealistic outputs with exceptional temporal and cross-view coherence.

Figures

Figures reproduced from arXiv: 2504.19614 by the authors.

Figure 1
Figure 1. DiVE in qualitative visualizations and comparison of results under different acceleration techniques. (a) DiVE generates [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of DiVE for multi-view video generation. Our model encodes four inputs for controllable generation: scene [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The overall process of Multi-Control Auxiliary [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (9 more)
Figure 5
Figure 5. Figure 5: Visualization of quantitative comparison. [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Qualitative comparison of DiVE with MagicDrive and Panacea. We use dashed boxes to highlight some of the [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 8
Figure 8. Figure 8: DiVE vs. W-CODA baseline [11], winners [8, 22, 34] and MagicDrive-V2 [10]. resource. Moreover, augmenting the original dataset with DiVE￾generated data further boosts StreamPETR’s performance across all metrics, with mAP and NDS gains exhibiting substantially larger ma…
Figure 9
Figure 9. Figure 9: Night scene generation in different CFG methods. [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: Effect of camera information. D More Visualization Results D.1 Flexible Controllable Generation [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 11
Figure 11. Figure 11: Flexible controllable generation from DiVE. [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]
Figure 12
Figure 12. Figure 12: Qualitative comparison with MagicDrive and Panacea on driving scene from nuScenes validation set. [PITH_FULL_IMAGE:figures/full_fig_p014_12.png]
Figure 13
Figure 13. Figure 13: Qualitative comparison with MagicDrive and Panacea on driving scene from nuScenes validation set. [PITH_FULL_IMAGE:figures/full_fig_p015_13.png]
Figure 14
Figure 14. Figure 14: Long video generated by DiVE [PITH_FULL_IMAGE:figures/full_fig_p016_14.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. A Survey of World Models for Autonomous Driving

    cs.RO 2025-01 conditional novelty 2.0 of 10

    A survey presenting a three-branch taxonomy of world models for autonomous driving, plus benchmark tables comparing representative generation and planning methods on nuScenes, Waymo, Occ3D, and CarlaSC.

Reference graph

Works this paper leans on

71 extracted references · 37 canonical work pages · cited by 1 Pith paper

  1. [1]

    Andreas Blattmann, Tim Dockhorn, Sumith Kulal, Daniel Mendelevitch, Maciej Kilian, Dominik Lorenz, Yam Levi, Zion English, Vikram Voleti, Adam Letts, et al. 2023. Stable video diffusion: Scaling latent video diffusion models to large datasets. arXiv preprint arXiv:2311.15127 (2023)

    arXiv 2023

  2. [2]

    Tim Brooks, Aleksander Holynski, and Alexei A Efros. 2023. Instructpix2pix: Learning to follow image editing instructions. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 18392–18402

    2023

  3. [3]

    Holger Caesar, Varun Bankiti, Alex H Lang, Sourabh Vora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom. 2020. nuscenes: A multimodal dataset for autonomous driving. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 11621–11631

    2020

  4. [4]

    Junsong Chen, YU Jincheng, GE Chongjian, Lewei Yao, Enze Xie, Zhongdao Wang, James Kwok, Ping Luo, Huchuan Lu, and Zhenguo Li. 2024. PixArt-𝛼: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis. In The Twelfth International Conference on Learning Representations

    work page 2024

  5. [5]

    Junsong Chen, Yue Wu, Simian Luo, Enze Xie, Sayak Paul, Ping Luo, Hang Zhao, and Zhenguo Li. 2024. PixArt-𝛿: Fast and Controllable Image Generation with Latent Consistency Models. In ICML 2024 Workshop on Theoretical Foundations of Foundation Models

    work page 2024

  6. [6]

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xi- aohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. 2021. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. InInternational Conference on Learning Representations

    2021

  7. [7]

    Ruoyi Du, Dongliang Chang, Timothy Hospedales, Yi-Zhe Song, and Zhanyu Ma. 2024. Demofusion: Democratising high-resolution image generation with no $$$. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 6159–6168

    work page 2024

  8. [8]

    Zhiying Du and Zhen Xing. 2024. Challenge report: Track 2 of Multimodal Perception and Comprehension of Corner Cases in Autonomous Driving. In ECCV 2024 Workshop on Multimodal Perception and Comprehension of Corner Cases in Autonomous Driving

    work page 2024

Show all 71 references
  1. [9]

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, et al. 2024. Scaling rectified flow transformers for high-resolution image synthesis. In Forty- first International Conference on M...

    2024

  2. [10]

    Ruiyuan Gao, Kai Chen, Bo Xiao, Lanqing Hong, Zhenguo Li, and Qiang Xu

  3. [11]

    Ruiyuan Gao, Kai Chen, Enze Xie, Hong Lanqing, Zhenguo Li, Dit-Yan Yeung, and Qiang Xu. 2024. MagicDrive: Street View Generation with Diverse 3D Geometry Control. In The Twelfth International Conference on Learning Representations

    2024

  4. [12]

    Srinivasan, Jonathan T

    Ruiqi Gao, Aleksander Holynski, Philipp Henzler, Arthur Brussee, Ricardo Martin Brualla, Pratul P. Srinivasan, Jonathan T. Barron, and Ben Poole. 2024. CAT3D: Create Anything in 3D with Multi-View Diffusion Models. In The Thirty-eighth Annual Conference on Neural Information P...

    2024

  5. [13]

    Lanqing Guo, Yingqing He, Haoxin Chen, Menghan Xia, Xiaodong Cun, Yufei Wang, Siyu Huang, Yong Zhang, Xintao Wang, Qifeng Chen, et al. 2024. Make a cheap scaling: A self-cascade diffusion model for higher-resolution adaptation. In European Conference on Computer Vision . Sprin...

    2024

  6. [14]

    Hao He, Ceyuan Yang, Shanchuan Lin, Yinghao Xu, Meng Wei, Liangke Gui, Qi Zhao, Gordon Wetzstein, Lu Jiang, and Hongsheng Li. 2025. CameraCtrl II: Dynamic Scene Exploration via Camera-controlled Video Diffusion Models. arXiv preprint arXiv:2503.10592 (2025)

    2025 arXiv

  7. [15]

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. 2017. Gans trained by a two time-scale update rule converge to a local nash equilibrium. Advances in neural information processing systems 30 (2017)

    2017

  8. [16]

    Jonathan Ho and Tim Salimans. 2021. Classifier-Free Diffusion Guidance. In NeurIPS 2021 Workshop on Deep Generative Models and Downstream Applications

    2021

  9. [17]

    Emiel Hoogeboom, Jonathan Heek, and Tim Salimans. 2023. simple diffusion: End-to-end diffusion for high resolution images. In International Conference on Machine Learning. PMLR, 13213–13232

    2023

  10. [18]

    Yi-Ting Hsiao, Siavash Khodadadeh, Kevin Duarte, Wei-An Lin, Hui Qu, Mingi Kwon, and Ratheesh Kalarot. 2024. Plug-and-play diffusion distillation. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 13743–13752

    2024

  11. [19]

    Junjie Huang and Guan Huang. 2022. Bevdet4d: Exploit temporal cues in multi- camera 3d object detection. arXiv preprint arXiv:2203.17054 (2022)

    2022 arXiv

  12. [20]

    Junjie Huang, Guan Huang, Zheng Zhu, Yun Ye, and Dalong Du. 2021. Bevdet: High-performance multi-camera 3d object detection in bird-eye-view. arXiv preprint arXiv:2112.11790 (2021)

    2021 arXiv

  13. [21]

    Cristian Perez Jensen and Seyedmorteza Sadat. 2025. Efficient Distillation of Classifier-Free Guidance using Adapters. arXiv preprint arXiv:2503.07274 (2025)

    2025 arXiv

  14. [22]

    Junpeng Jiang, Gangyi Hong, Lijun Zhou, Enhui Ma, Hengtong Hu, xia zhou, Jie Xiang, Fan Liu, Kaicheng Yu, Haiyang Sun, Kun Zhan, Peng Jia, and Miao Zhang. 2024. DiVE: DiT-based Video Generation with Enhanced Control. In ECCV 2024 Workshop on Multimodal Perception and Comprehen...

    2024

  15. [23]

    Nikolai Kalischek, Michael Oechsle, Fabian Manhardt, Philipp Henzler, Konrad Schindler, and Federico Tombari. 2025. CubeDiff: Repurposing Diffusion-Based Image Models for Panorama Generation. In The Thirteenth International Confer- ence on Learning Representations

    2025

  16. [24]

    Bing Li, Cheng Zheng, Wenxuan Zhu, Jinjie Mai, Biao Zhang, Peter Wonka, and Bernard Ghanem. 2024. Vivid-ZOO: Multi-View Video Generation with Diffusion Model. In The Thirty-eighth Annual Conference on Neural Information Processing Systems

    2024

  17. [25]

    Renjie Li, Panwang Pan, Bangbang Yang, Dejia Xu, Shijie Zhou, Xuanyang Zhang, Zeming Li, Achuta Kadambi, Zhangyang Wang, Zhengzhong Tu, and Zhiwen Fan

  18. [26]

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

    2023

  19. [27]

    Zhiqi Li, Wenhai Wang, Hongyang Li, Enze Xie, Chonghao Sima, Tong Lu, Yu Qiao, and Jifeng Dai. 2022. Bevformer: Learning bird’s-eye-view representa- tion from multi-camera images via spatiotemporal transformers. In European conference on computer vision . Springer, 1–18

    2022

  20. [28]

    Buyu Liu, Kai Wang, Yansong Liu, Jun Bao, Tingting Han, and Jun Yu. 2024. MVPbev: Multi-view Perspective Image Generation from BEV with Test-time Controllability and Generalizability. In ACM Multimedia 2024

    2024

  21. [29]

    Gongye Liu, Menghan Xia, Yong Zhang, Haoxin Chen, Jinbo Xing, Yibo Wang, Xintao Wang, Ying Shan, and Yujiu Yang. 2024. StyleCrafter: Taming Artistic Video Diffusion with Reference-Augmented Adapter Learning.ACM Transactions on Graphics (TOG) 43, 6 (2024), 1–10

    2024

  22. [30]

    Haisong Liu, Yao Teng, Tao Lu, Haiguang Wang, and Limin Wang. 2023. Sparse- bev: High-performance sparse 3d object detection from multi-camera videos. In Proceedings of the IEEE/CVF International Conference on Computer Vision . 18580– 18590

    2023

  23. [31]

    Xingchao Liu, Chengyue Gong, and Qiang Liu. 2023. Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow. In The Eleventh International Conference on Learning Representations

    2023

  24. [32]

    Zhijian Liu, Haotian Tang, Alexander Amini, Xinyu Yang, Huizi Mao, Daniela L Rus, and Song Han. 2023. Bevfusion: Multi-task multi-sensor fusion with unified bird’s-eye view representation. In 2023 IEEE international conference on robotics and automation (ICRA). IEEE, 2774–2781

    2023

  25. [33]

    Nanye Ma, Mark Goldstein, Michael S Albergo, Nicholas M Boffi, Eric Vanden- Eijnden, and Saining Xie. 2024. Sit: Exploring flow and diffusion-based generative models with scalable interpolant transformers. In European Conference on Com- puter Vision. Springer, 23–40

    2024

  26. [34]

    Jianbiao Mei, Yukai Ma, Xuemeng Yang, Licheng Wen, Tiantian Wei, Min Dou, Botian Shi, and Yong Liu. 2024. DreamForge: Motion-Aware Autoregressive Video Generation for Multi-View Driving Scenes. In ECCV 2024 Workshop on Multimodal Perception and Comprehension of Corner Cases in...

    2024

  27. [35]

    Chenlin Meng, Robin Rombach, Ruiqi Gao, Diederik Kingma, Stefano Ermon, Jonathan Ho, and Tim Salimans. 2023. On distillation of guided diffusion mod- els. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 14297–14306

    2023

  28. [36]

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. 2021. Nerf: Representing scenes as neural radiance fields for view synthesis. Commun. ACM 65, 1 (2021), 99–106

    2021

  29. [37]

    OpenAI. 2024. Sora. (2024). https://openai.com/index/sora/

    2024

  30. [38]

    William Peebles and Saining Xie. 2023. Scalable diffusion models with transform- ers. In Proceedings of the IEEE/CVF International Conference on Computer Vision . 4195–4205

    2023

  31. [39]

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas Müller, Joe Penna, and Robin Rombach. 2024. SDXL: Improving Latent Dif- fusion Models for High-Resolution Image Synthesis. In The Twelfth International Conference on Learning Representations

    2024

  32. [40]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. 2021. Learning transferable visual models from natural language supervision. In International conference on machine learni...

    2021

  33. [41]

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, et al. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of machine learning research 21, 140 (2020), 1–67

    2020

  34. [42]

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 10684–10695

    2022

  35. [43]

    Jiaming Sun, Zehong Shen, Yuang Wang, Hujun Bao, and Xiaowei Zhou. 2021. LoFTR: Detector-free local feature matching with transformers. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 8922–8931. Preprint, Under review, April 2025 Junpeng ...

    2021

  36. [44]

    Alexander Swerdlow, Runsheng Xu, and Bolei Zhou. 2024. Street-view image generation from a bird’s-eye view layout. IEEE Robotics and Automation Letters (2024)

    2024

  37. [45]

    Thomas Unterthiner, Sjoerd Van Steenkiste, Karol Kurach, Raphael Marinier, Marcin Michalski, and Sylvain Gelly. 2018. Towards accurate generative models of video: A new metric & challenges. arXiv preprint arXiv:1812.01717 (2018)

    2018 arXiv

  38. [46]

    Qian Wang, Weiqi Li, Chong Mou, Xinhua Cheng, and Jian Zhang. 2024. 360dvd: Controllable panorama video generation with 360-degree video diffusion model. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recog- nition. 6913–6923

    2024

  39. [47]

    Shihao Wang, Yingfei Liu, Tiancai Wang, Ying Li, and Xiangyu Zhang. 2023. Exploring object-centric temporal modeling for efficient multi-view 3d object detection. In Proceedings of the IEEE/CVF International Conference on Computer Vision. 3621–3631

    2023

  40. [48]

    Xiaofeng Wang, Zheng Zhu, Guan Huang, Xinze Chen, Jiagang Zhu, and Jiwen Lu

  41. [49]

    Yuqi Wang, Jiawei He, Lue Fan, Hongxin Li, Yuntao Chen, and Zhaoxiang Zhang

  42. [50]

    WanTeam, Ang Wang, Baole Ai, Bin Wen, Chaojie Mao, Chen-Wei Xie, Di Chen, Feiwu Yu, Haiming Zhao, Jianxiao Yang, Jianyuan Zeng, Jiayu Wang, Jingfeng Zhang, Jingren Zhou, Jinkai Wang, Jixuan Chen, Kai Zhu, Kang Zhao, Keyu Yan, Lianghua Huang, Mengyang Feng, Ningyi Zhang, Panden...

    2025 arXiv

  43. [51]

    In European Conference on Computer Vision

    DriveDreamer: Towards Real-World-Drive World Models for Autonomous Driving. In European Conference on Computer Vision . Springer, 55–72

  44. [52]

    Yuqing Wen, Yucheng Zhao, Yingfei Liu, Binyuan Huang, Fan Jia, Yanhui Wang, Chi Zhang, Tiancai Wang, Xiaoyan Sun, and Xiangyu Zhang. 2023. Panacea+: Panoramic and Controllable Video Generation for Autonomous Driving. arXiv preprint arXiv:2408.07605 (2023)

    2023 arXiv

  45. [53]

    In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Driving into the Future: Multiview Visual Forecasting and Planning with World Model for Autonomous Driving. In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

  46. [54]

    Haoning Wu, Shaocheng Shen, Qiang Hu, Xiaoyun Zhang, Ya Zhang, and Yanfeng Wang. 2025. MegaFusion: Extend Diffusion Models towards Higher-resolution Image Generation without Further Tuning. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (W ACV)

    2025

  47. [55]

    Zijian Zhang Rox Min Zuozhuo Dai Jin Zhou Jiangfeng Xiong Xin Li Bo Wu Jianwei Zhang Kathrina Wu Qin Lin Aladdin Wang Andong Wang Changlin Li Duojun Huang Fang Yang Hao Tan Hongmei Wang Jacob Song Jiawang Bai Jianbing Wu Jinbao Xue Joey Wang Junkun Yuan Kai Wang Mengyang Liu P...

  48. [56]

    https://arxiv.org/abs/2412.03603

    HunyuanVideo: A Systematic Framework For Large Video Generative Models. https://arxiv.org/abs/2412.03603

    arXiv

  49. [57]

    Chenyu Yang, Yuntao Chen, Hao Tian, Chenxin Tao, Xizhou Zhu, Zhaoxiang Zhang, Gao Huang, Hongyang Li, Yu Qiao, Lewei Lu, et al. 2023. Bevformer v2: Adapting modern image backbones to bird’s-eye-view recognition via perspective supervision. In Proceedings of the IEEE/CVF Confer...

    2023

  50. [58]

    Yuqing Wen, Yucheng Zhao, Yingfei Liu, Fan Jia, Yanhui Wang, Chong Luo, Chi Zhang, Tiancai Wang, Xiaoyan Sun, and Xiangyu Zhang. 2024. Panacea: Panoramic and controllable video generation for autonomous driving. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision an...

    2024

  51. [59]

    Zhuoyi Yang, Jiayan Teng, Wendi Zheng, Ming Ding, Shiyu Huang, Jiazheng Xu, Yuanming Yang, Wenyi Hong, Xiaohan Zhang, Guanyu Feng, Da Yin, Yux- uan.Zhang, Weihan Wang, Yean Cheng, Bin Xu, Xiaotao Gu, Yuxiao Dong, and Jie Tang. 2025. CogVideoX: Text-to-Video Diffusion Models wi...

    2025

  52. [60]

    Bin Xie, Yingfei Liu, Tiancai Wang, Jiale Cao, and Xiangyu Zhang. 2025. Glad: A Streaming Scene Generator for Autonomous Driving. In The Thirteenth Interna- tional Conference on Learning Representations

    2025

  53. [61]

    Yiming Xie, Chun-Han Yao, Vikram Voleti, Huaizu Jiang, and Varun Jampani

  54. [62]

    In The Thirteenth International Conference on Learning Representa- tions

    SV4D: Dynamic 3D Content Generation with Multi-Frame and Multi-View Consistency. In The Thirteenth International Conference on Learning Representa- tions

  55. [63]

    Zangwei Zheng, Xiangyu Peng, Tianji Yang, Chenhui Shen, Shenggui Li, Hongxin Liu, Yukun Zhou, Tianyi Li, and Yang You. 2024. Open-sora: Democratizing efficient video production for all. arXiv preprint arXiv:2412.20404 (2024)

    2024 arXiv

  56. [64]

    Kairui Yang, Enhui Ma, Jibin Peng, Qing Guo, Di Lin, and Kaicheng Yu. 2023. Bevcontrol: Accurately controlling street-view elements with multi-perspective consistency via bev sketch layout. arXiv preprint arXiv:2308.01661 (2023)

    2023 arXiv

  57. [65]

    Zhenyu Zhou, Defang Chen, Can Wang, Chun Chen, and Siwei Lyu. 2025. DICE: Distilling Classifier-Free Guidance into Text Embeddings. arXiv preprint arXiv:2502.03726 (2025). DiVE: Efficient Multi-View Driving Scenes Generation Based on Video Diffusion Transformer Preprint, Under...

    2025

  58. [66]

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. 2023. Adding conditional con- trol to text-to-image diffusion models. InProceedings of the IEEE/CVF international conference on computer vision . 3836–3847

    2023

  59. [67]

    Guosheng Zhao, Xiaofeng Wang, Zheng Zhu, Xinze Chen, Guan Huang, Xiaoyi Bao, and Xingang Wang. 2024. Drivedreamer-2: Llm-enhanced world models for diverse driving video generation. arXiv preprint arXiv:2403.06845 (2024)

    2024 arXiv

  60. [68]

    Guangcong Zheng, Xianpan Zhou, Xuewei Li, Zhongang Qi, Ying Shan, and Xi Li. 2023. Layoutdiffusion: Controllable diffusion model for layout-to-image generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 22490–22499

    2023

  61. [70]

    Brady Zhou and Philipp Krähenbühl. 2022. Cross-view transformers for real-time map-view semantic segmentation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 13760–13769

    2022

  62. [2024]

    arXiv preprint arXiv:2411.13807 (2024)

    MagicDrive-V2: High-Resolution Long Video Generation for Autonomous Driving with Adaptive Control. arXiv preprint arXiv:2411.13807 (2024)

    2024 arXiv

  63. [2025]

    In The Thirteenth International Conference on Learning Representations

    4K4DGen: Panoramic 4D Generation at 4K Resolution. In The Thirteenth International Conference on Learning Representations

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.