pith. machine review for the scientific record. sign in

arxiv: 2604.18468 · v1 · submitted 2026-04-20 · 💻 cs.CV · cs.AI· cs.GR· cs.LG

Recognition: unknown

Asset Harvester: Extracting 3D Assets from Autonomous Driving Logs for Simulation

Tianshi Cao , Jiawei Ren , Yuxuan Zhang , Jaewoo Seo , Jiahui Huang , Shikhar Solanki , Haotian Zhang , Mingfei Guo
show 7 more authors
Haithem Turki Muxingzi Li Yue Zhu Sipeng Zhang Zan Gojcic Sanja Fidler Kangxue Yin
Authors on Pith no claims yet

Pith reviewed 2026-05-10 04:14 UTC · model grok-4.3

classification 💻 cs.CV cs.AIcs.GRcs.LG
keywords 3D asset extractionautonomous driving logssparse view reconstructionmultiview generation3D Gaussian liftingsimulation assetsimage-to-3D pipeline
0
0 comments X

The pith

Asset Harvester converts sparse object views from driving logs into complete 3D assets for simulation.

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

The paper introduces an image-to-3D pipeline that takes limited camera observations of objects captured during real autonomous vehicle drives and turns them into full, reusable 3D models. It achieves this through large-scale collection of object-centric data from logs, geometry-aware cleaning across different sensors, and a model called SparseViewDiT that first generates missing views from sparse inputs before lifting the results into 3D Gaussian representations. If the approach works, it fills a gap in closed-loop simulation where scene-level reconstruction alone cannot supply manipulable objects for testing vehicle interactions and safety scenarios. The system is presented as a complete end-to-end solution rather than a single isolated component.

Core claim

Asset Harvester is an image-to-3D model and end-to-end pipeline that converts sparse, in-the-wild object observations from real driving logs into complete, simulation-ready assets through large-scale curation of object-centric training tuples, geometry-aware preprocessing across heterogeneous sensors, and a training recipe that couples sparse-view-conditioned multiview generation with 3D Gaussian lifting inside SparseViewDiT.

What carries the argument

SparseViewDiT, a diffusion transformer that performs sparse-view-conditioned multiview generation and then applies 3D Gaussian lifting to produce complete object geometry from limited-angle inputs.

If this is right

  • Simulation environments gain individual, manipulable 3D objects that can be moved or interacted with independently of the background scene.
  • Novel-view synthesis becomes possible at large viewpoint changes that scene reconstruction methods cannot handle.
  • Real observed objects from driving logs become reusable assets instead of requiring manual modeling or synthetic generation.
  • Closed-loop testing scales by automatically populating simulations with objects encountered in actual AV data.

Where Pith is reading between the lines

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

  • The resulting assets could be inserted into existing simulation frameworks to test rare or dangerous object interactions without new data collection.
  • The same curation-plus-lifting recipe might apply to other domains that collect sparse object views, such as indoor robotics.
  • Combining these object assets with existing neural scene reconstructions could produce fully interactive driving environments.

Load-bearing premise

That real-world driving log data, once curated and preprocessed, supplies enough signal for the model to reliably fill in missing geometry and appearances from limited viewing angles.

What would settle it

Extract assets from a held-out set of driving log observations and measure whether the resulting 3D models produce view-consistent renderings or match independent high-quality scans when viewed from angles absent in the original logs.

Figures

Figures reproduced from arXiv: 2604.18468 by Haithem Turki, Haotian Zhang, Jaewoo Seo, Jiahui Huang, Jiawei Ren, Kangxue Yin, Mingfei Guo, Muxingzi Li, Sanja Fidler, Shikhar Solanki, Sipeng Zhang, Tianshi Cao, Yue Zhu, Yuxuan Zhang, Zan Gojcic.

Figure 1
Figure 1. Figure 1: Overview of Asset Harvester. Starting from large-scale AV logs stored in NCore, we crop and rectify object observations, generate multiview images with SparseViewDiT, lift them into 3D assets with an Object TokenGS, and reinsert the assets into scenes with harmonization for closed-loop simulation. addition, supervisory signals such as 3D cuboid tracks can be noisy or temporally unstable. Camera calibration… view at source ↗
Figure 2
Figure 2. Figure 2: Architecture overview of SparseViewDiT for sparse-view-conditioned multi-view generation. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: In-the-wild qualitative results across diverse object classes, including sedan, bus, trailer, trash bin, [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative comparison against image-to-3D baselines (zoom in for details). Asset Harvester is [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Input-view ablation for Asset Harvester. We compare 1-view input results against 4-view input results. While single-view input already yields plausible reconstructions, additional input views with better object coverage improve sharpness and details. Inference speed We measure the end-to-end inference time of Asset Harvester on single NVIDIA A100 and H100 GPU and report the results in [PITH_FULL_IMAGE:fig… view at source ↗
Figure 6
Figure 6. Figure 6: OOD image editing and generalization with [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Pedestrian animation results with Asset Harvester. We convert input observations into an A-pose asset, rig the generated asset with a simple LBS implemented with SOMA and GEM, and animate it with Kimodo. Pedestrian Animation For pedestrian animation, we first take the first input view and convert it to an A-pose image using Qwen￾Image-Edit-2511 [24] with the prompt shown below. Change the posture of the pe… view at source ↗
Figure 8
Figure 8. Figure 8: Insertion and harmonization results with [PITH_FULL_IMAGE:figures/full_fig_p016_8.png] view at source ↗
read the original abstract

Closed-loop simulation is a core component of autonomous vehicle (AV) development, enabling scalable testing, training, and safety validation before real-world deployment. Neural scene reconstruction converts driving logs into interactive 3D environments for simulation, but it does not produce complete 3D object assets required for agent manipulation and large-viewpoint novel-view synthesis. To address this challenge, we present Asset Harvester, an image-to-3D model and end-to-end pipeline that converts sparse, in-the-wild object observations from real driving logs into complete, simulation-ready assets. Rather than relying on a single model component, we developed a system-level design for real-world AV data that combines large-scale curation of object-centric training tuples, geometry-aware preprocessing across heterogeneous sensors, and a robust training recipe that couples sparse-view-conditioned multiview generation with 3D Gaussian lifting. Within this system, SparseViewDiT is explicitly designed to address limited-angle views and other real-world data challenges. Together with hybrid data curation, augmentation, and self-distillation, this system enables scalable conversion of sparse AV object observations into reusable 3D assets.

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

1 major / 1 minor

Summary. The paper introduces Asset Harvester, an image-to-3D model and end-to-end pipeline that converts sparse, in-the-wild object observations from real autonomous driving logs into complete, simulation-ready 3D assets. It combines large-scale curation of object-centric training tuples, geometry-aware preprocessing across heterogeneous sensors, and a training recipe coupling sparse-view-conditioned multiview generation with 3D Gaussian lifting in a model called SparseViewDiT, augmented by hybrid data curation, augmentation, and self-distillation to handle limited-angle views and other real-world challenges.

Significance. If the central claims hold with supporting evidence, the work would address a notable gap in neural scene reconstruction for AV simulation by producing reusable, manipulable 3D object assets suitable for agent manipulation and wide-baseline novel-view synthesis, potentially improving the scalability and realism of closed-loop testing and safety validation.

major comments (1)
  1. [Abstract] Abstract: The manuscript asserts that the system-level design 'enables scalable conversion of sparse AV object observations into reusable 3D assets,' yet the provided description supplies no quantitative results, error metrics (e.g., completeness, PSNR/SSIM for novel views, or mesh quality), ablation studies isolating components such as SparseViewDiT or self-distillation, or comparisons against baselines on real or held-out AV data. This absence leaves the claim that the recipe overcomes the ill-posedness of inferring unseen geometry from limited-angle, noisy observations untested and load-bearing for the central contribution.
minor comments (1)
  1. [Abstract] Abstract: The term 'SparseViewDiT' is introduced as 'explicitly designed' without a brief definition or reference to its architecture details at first mention, which could be clarified for readers unfamiliar with the DiT backbone.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback and for recognizing the potential impact of Asset Harvester on scalable 3D asset creation for AV simulation. We address the major comment point-by-point below and will incorporate revisions to strengthen the presentation of our empirical results.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The manuscript asserts that the system-level design 'enables scalable conversion of sparse AV object observations into reusable 3D assets,' yet the provided description supplies no quantitative results, error metrics (e.g., completeness, PSNR/SSIM for novel views, or mesh quality), ablation studies isolating components such as SparseViewDiT or self-distillation, or comparisons against baselines on real or held-out AV data. This absence leaves the claim that the recipe overcomes the ill-posedness of inferring unseen geometry from limited-angle, noisy observations untested and load-bearing for the central contribution.

    Authors: We agree that the abstract, in its current form, does not explicitly summarize the quantitative evidence supporting the central claims. The full manuscript contains these evaluations in the Experiments section, including PSNR/SSIM and completeness metrics for novel-view synthesis and 3D asset quality, ablation studies isolating SparseViewDiT, hybrid curation, augmentation, and self-distillation, as well as comparisons to baselines on both synthetic and real held-out AV logs. To directly address the referee's concern, we will revise the abstract to concisely incorporate key quantitative highlights from these results, making the empirical support for overcoming limited-angle and noisy observations explicit at the summary level. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical system description only

full rationale

The paper presents Asset Harvester as an end-to-end pipeline combining large-scale object-centric curation, geometry-aware preprocessing, and a SparseViewDiT training recipe (sparse-view multiview generation plus 3D Gaussian lifting) with augmentation and self-distillation. No equations, fitted parameters, predictions, or first-principles derivations appear; the central claim is a system-level empirical design rather than a mathematical reduction. No self-citation chains, ansatzes, or uniqueness theorems are invoked as load-bearing steps in the provided text. The derivation chain is therefore self-contained and non-circular.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

The abstract introduces a new model and pipeline but contains no explicit mathematical assumptions, free parameters, or invented physical entities; all components are described at the system level.

invented entities (1)
  • SparseViewDiT no independent evidence
    purpose: Address limited-angle views and real-world AV data challenges in multiview generation
    Presented as a custom architecture within the pipeline; no independent evidence of its performance outside this system is provided.

pith-pipeline@v0.9.0 · 5563 in / 1281 out tokens · 34819 ms · 2026-05-10T04:14:46.044291+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

53 extracted references · 23 canonical work pages · 5 internal anchors

  1. [1]

    Driving into the future: Multiview visual forecasting and planning with world model for autonomous driving, 2023

    Yuqi Wang, Jiawei He, Lue Fan, Hongxin Li, Yuntao Chen, and Zhaoxiang Zhang. Driving into the future: Multiview visual forecasting and planning with world model for autonomous driving, 2023. URLhttps://arxiv.org/abs/ 2311.17918. 1

    work page arXiv 2023

  2. [2]

    Understanding world or predicting future? a comprehensive survey of world models.ACM Computing Surveys, 58(3):1–38, 2025

    Jingtao Ding, Yunke Zhang, Yu Shang, Yuheng Zhang, Zefang Zong, Jie Feng, Yuan Yuan, Hongyuan Su, Nian Li, Nicholas Sukiennik, et al. Understanding world or predicting future? a comprehensive survey of world models.ACM Computing Surveys, 58(3):1–38, 2025. 1

    2025

  3. [3]

    Gaia-2: A controllable multi-view generative world model for autonomous driving.arXiv preprint arXiv:2503.20523, 2025

    Lloyd Russell, Anthony Hu, Lorenzo Bertoni, George Fedoseev, Jamie Shotton, Elahe Arani, and Gianluca Corrado. Gaia-2: A controllable multi-view generative world model for autonomous driving, 2025. URLhttps://arxiv.org/ abs/2503.20523. 1

    work page arXiv 2025

  4. [4]

    NVIDIA. Nurec. Website, 2026. URLhttps://research.nvidia.com/labs/sil/nurec/. Accessed 2026-03-08. 1, 17

    2026

  5. [5]

    3dgut: Enabling distorted cameras and secondary rays in gaussian splatting.Conference on Computer Vision and Pattern Recognition ( CVPR),

    Qi Wu, Janick Martinez Esturo, Ashkan Mirzaei, Nicolas Moenne-Loccoz, and Zan Gojcic. 3dgut: Enabling distorted cameras and secondary rays in gaussian splatting.Conference on Computer Vision and Pattern Recognition ( CVPR),

  6. [6]

    SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformers

    Enze Xie, Junsong Chen, Junyu Chen, Han Cai, Haotian Tang, Yujun Lin, Zhekai Zhang, Muyang Li, Ligeng Zhu, Yao Lu, and Song Han. Sana: Efficient high-resolution image synthesis with linear diffusion transformers, 2024. URL https://arxiv.org/abs/2410.10629. 2, 4, 5

    work page internal anchor Pith review arXiv 2024

  7. [7]

    Tokengs: Decoupling 3d gaussian prediction from pixels with learnable tokens

    Jiawei Ren, Michal Tyszkiewicz, Jiahui Huang, and Zan Gojcic. Tokengs: Decoupling 3d gaussian prediction from pixels with learnable tokens. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition ( CVPR), 2026. 2, 5

    2026

  8. [8]

    NVIDIA. Ncore. GitHub repository, 2026. URLhttps://github.com/NVIDIA/ncore. Accessed 2026-03-08. 2, 3

    2026

  9. [9]

    Diffusionharmonizer: Bridging neural reconstruction and photorealistic simulation with online diffusion enhancer.arXiv preprint arXiv:2602.24096, 2026

    YuxuanZhang, KatarínaTóthová, ZianWang, KangxueYin, HaithemTurki, RiccardodeLutio, Yen-YuChang, OrLitany, Sanja Fidler, and Zan Gojcic. Diffusionharmonizer: Bridging neural reconstruction and photorealistic simulation with online diffusion enhancer.arXiv preprint arXiv:2602.24096, 2026. URLhttps://arxiv.org/abs/2602.24096. 2, 15

    work page arXiv 2026

  10. [10]

    Juho Kannala and Sami S Brandt. A generic camera model and calibration method for conventional, wide-angle, and fish-eye lenses.IEEE transactions on pattern analysis and machine intelligence, 28(8):1335–1340, 2006. 3

    2006

  11. [11]

    Schwing, and Alexander Kirillov

    Bowen Cheng, Alexander G. Schwing, and Alexander Kirillov. Per-pixel classification is not all you need for semantic segmentation. 2021. 3

    2021

  12. [12]

    Flow Matching for Generative Modeling

    Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximilian Nickel, and Matt Le. Flow matching for generative modeling.arXiv preprint arXiv:2210.02747, 2022. 3

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  13. [13]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023. 4

    2023

  14. [14]

    Deep compression autoencoder for efficient high-resolution diffusion models

    Junyu Chen, Han Cai, Junsong Chen, Enze Xie, Shang Yang, Haotian Tang, Muyang Li, Yao Lu, and Song Han. Deep compression autoencoder for efficient high-resolution diffusion models.arXiv preprint arXiv:2410.10733, 2024. 4

    work page arXiv 2024

  15. [15]

    NVIDIA. C-radio. Hugging Face model card, 2026. URLhttps://huggingface.co/nvidia/C-RADIO. Accessed 2026-03-09. 5

    2026

  16. [16]

    Srinivasan, Matthew Tancik, Jonathan T

    Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view synthesis. InEuropean Conference on Computer Vision (ECCV),

  17. [17]

    An image is worth 16x16 words: Transformers for image recognition at scale.ICLR, 2021

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale.ICLR, 2021. 6 19 Asset Harvester: Extracting 3D Assets from Autonomous Dr...

    2021

  18. [18]

    Lgm: Large multi-view gaussian model for high-resolution 3d content creation.arXiv preprint arXiv:2402.05054, 2024

    Jiaxiang Tang, Zhaoxi Chen, Xiaokang Chen, Tengfei Wang, Gang Zeng, and Ziwei Liu. Lgm: Large multi-view gaussian model for high-resolution 3d content creation.arXiv preprint arXiv:2402.05054, 2024. 6

    work page arXiv 2024

  19. [19]

    Objaverse: A universe of annotated 3d objects

    Matt Deitke, Dustin Schwenk, Jordi Salvador, Luca Weihs, Oscar Michel, Eli VanderBilt, Ludwig Schmidt, Kiana Ehsani, Aniruddha Kembhavi, and Ali Farhadi. Objaverse: A universe of annotated 3d objects.arXiv preprint arXiv:2212.08051,

    work page arXiv

  20. [20]

    Qwen2.5-vl, January 2025

    Qwen Team. Qwen2.5-vl, January 2025. URLhttps://qwenlm.github.io/blog/qwen2.5-vl/. 7

    2025

  21. [21]

    Nvidia omniverse replicator

    NVIDIA. Nvidia omniverse replicator. Website, 2026. URL https://developer.nvidia.com/omniverse/ replicator. Accessed 2026-02-27. 7

    2026

  22. [22]

    Renderpeople

    RenderPeople. Renderpeople. Website, 2026. URLhttps://renderpeople.com/. Accessed 2026-02-27. 7

    2026

  23. [23]

    Polyhaven

    Poly Haven. Polyhaven. Website, 2026. URLhttps://polyhaven.com/. Accessed 2026-02-27. 7

    2026

  24. [24]

    Qwen-Image Technical Report

    Chenfei Wu, Jiahao Li, Jingren Zhou, Junyang Lin, Kaiyuan Gao, Kun Yan, Sheng ming Yin, Shuai Bai, Xiao Xu, Yilei Chen, Yuxiang Chen, Zecheng Tang, Zekai Zhang, Zhengyi Wang, An Yang, Bowen Yu, Chen Cheng, Dayiheng Liu, Deqing Li, Hang Zhang, Hao Meng, Hu Wei, Jingyuan Ni, Kai Chen, Kuan Cao, Liang Peng, Lin Qu, Minggang Wu, Peng Wang, Shuting Yu, Tingkun...

    work page internal anchor Pith review arXiv 2025

  25. [25]

    OpenAI. Gpt-5.2. OpenAI, 2026. URLhttps://openai.com/index/introducing-gpt-5-2/. Accessed 2026-03-26. 9

    2026

  26. [26]

    Oriane Simeoni, Huy V. Vo, Maximilian Seitzer, Federico Baldassarre, Maxime Oquab, Cijo Jose, Vasil Khalidov, Marc Szafraniec, Seungeun Yi, Michaël Ramamonjisoa, Francisco Massa, Daniel Haziza, Luca Wehrstedt, Jianyuan Wang, Timothée Darcet, Théo Moutakanni, Leonel Sentana, Claire Roberts, Andrea Vedaldi, Jamie Tolan, John Brandt, Camille Couprie, Julien ...

    2025

  27. [27]

    Sam 3d body: Robust full-body human mesh recovery.arXiv preprint arXiv:2602.15989,

    Xitong Yang, Devansh Kukreja, Don Pinkus, Anushka Sagar, Taosha Fan, Jinhyung Park, Soyong Shin, Jinkun Cao, Jiawei Liu, Nicolas Ugrinovic, Matt Feiszli, Jitendra Malik, Piotr Dollar, and Kris Kitani. Sam 3d body: Robust full-body human mesh recovery.arXiv preprint arXiv:2602.15989, 2026. 10

    work page arXiv 2026

  28. [28]

    SAM 3D: 3Dfy Anything in Images

    SAM 3D Team, Xingyu Chen, Fu-Jen Chu, Pierre Gleize, Kevin J Liang, Alexander Sax, Hao Tang, Weiyao Wang, Michelle Guo, Thibaut Hardin, Xiang Li, Aohan Lin, Jiawei Liu, Ziqi Ma, Anushka Sagar, Bowen Song, Xiaodong Wang, Jianing Yang, Bowen Zhang, Piotr Dollár, Georgia Gkioxari, Matt Feiszli, and Jitendra Malik. Sam 3d: 3dfy anything in images, 2025. URLht...

    work page internal anchor Pith review arXiv 2025

  29. [29]

    arXiv preprint arXiv:2412.01506 (2024) 4

    Jianfeng Xiang, Zelong Lv, Sicheng Xu, Yu Deng, Ruicheng Wang, Bowen Zhang, Dong Chen, Xin Tong, and Jiaolong Yang. Structured 3d latents for scalable and versatile 3d generation.arXiv preprint arXiv:2412.01506, 2024. 11, 15

    work page arXiv 2024

  30. [30]

    Hunyuan3D 2.0: Scaling Diffusion Models for High Resolution Textured 3D Assets Generation

    Zibo Zhao, Zeqiang Lai, Qingxiang Lin, et al. Hunyuan3d 2.0: Scaling diffusion models for high resolution textured 3d assets generation.arXiv preprint arXiv:2501.12202, 2025. 11, 15

    work page Pith review arXiv 2025

  31. [31]

    Hunyuan3d 2.1: From images to high-fidelity 3d assets with production-ready pbr material, 2025

    Tencent Hunyuan3D Team. Hunyuan3d 2.1: From images to high-fidelity 3d assets with production-ready pbr material, 2025. 11

    2025

  32. [32]

    Gemini 3 flash (nano banana 2)

    Google. Gemini 3 flash (nano banana 2). [Large language model], 2026. URLhttps://gemini.google.com/. Accessed 2026-03-18. 14

    2026

  33. [33]

    Soma: Unifying parametric human body models.arXiv preprint arXiv:2603.16858, 2026

    Jun Saito, Jiefeng Li, Michael de Ruyter, Miguel Guerrero, Edy Lim, Ehsan Hassani, Roger Blanco Ribera, Hyejin Moon, Magdalena Dadela, Marco Di Lucca, Qiao Wang, Xueting Li, Jan Kautz, Simon Yuen, and Umar Iqbal. Soma: Unifying parametric human body models.arXiv preprint arXiv:2603.16858, 2026. URLhttps://arxiv.org/abs/2603.16858. 15

    work page arXiv 2026

  34. [34]

    Genmo: A generalist model for human motion

    Jiefeng Li, Jinkun Cao, Haotian Zhang, Davis Rempe, Jan Kautz, Umar Iqbal, and Ye Yuan. Genmo: A generalist model for human motion. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2025. 15 20 Asset Harvester: Extracting 3D Assets from Autonomous Driving Logs for Simulation

    2025

  35. [35]

    Kimodo: Scaling controllable human motion generation.arXiv, 2026

    Davis Rempe, Mathis Petrovich, Ye Yuan, Haotian Zhang, Xue Bin Peng, Yifeng Jiang, Tingwu Wang, Umar Iqbal, David Minor, Michael de Ruyter, Jiefeng Li, Chen Tessler, Edy Lim, Eugene Jeong, Sam Wu, Ehsan Hassani, Michael Huang, Jin-Bey Yu, Chaeyeon Chung, Lina Song, Olivier Dionne, Jan Kautz, Simon Yuen, and Sanja Fidler. Kimodo: Scaling controllable human...

    2026

  36. [36]

    Hunyuan3d 1.0: A unified framework for text-to-3d and image-to-3d generation, 2024

    Tencent Hunyuan3D Team. Hunyuan3d 1.0: A unified framework for text-to-3d and image-to-3d generation, 2024. 15

    2024

  37. [37]

    arXiv preprint arXiv:2506.16504 , year=

    Tencent Hunyuan3D Team. Hunyuan3d 2.5: Towards high-fidelity 3d assets generation with ultimate details, 2025. URLhttps://arxiv.org/abs/2506.16504. 15

    work page arXiv 2025

  38. [38]

    Native and compact structured latents for 3d generation.arXiv preprint arXiv:2512.14692, 2025

    Jianfeng Xiang, Xiaoxue Chen, Sicheng Xu, Ruicheng Wang, Zelong Lv, Yu Deng, Hongyuan Zhu, Yue Dong, Hao Zhao, Nicholas Jing Yuan, and Jiaolong Yang. Native and compact structured latents for 3d generation.arXiv preprint arXiv:2512.14692, 2025. 15

    work page arXiv 2025

  39. [39]

    arXiv:2303.11328 , year=

    Ruoshi Liu, Rundi Wu, Basile Van Hoorick, Pavel Tokmakov, Sergey Zakharov, and Carl Vondrick. Zero-1-to-3: Zero-shot one image to 3d object.arXiv preprint arXiv:2303.11328, 2023. 16

    work page arXiv 2023

  40. [40]

    MVDream: Multi-view Diffusion for 3D Generation

    Yichun Shi, Peng Wang, Jianglong Ye, Mai Long, Kejie Li, and Xiao Yang. Mvdream: Multi-view diffusion for 3d generation.arXiv preprint arXiv:2308.16512, 2023. 16

    work page internal anchor Pith review arXiv 2023

  41. [41]

    arXiv preprint arXiv:2312.02201 , year=

    Peng Wang and Yichun Shi. Imagedream: Image-prompt multi-view diffusion for 3d generation.arXiv preprint arXiv:2312.02201, 2023. 16

    work page arXiv 2023

  42. [42]

    Direct3d: Scalable image-to-3d generation via 3d latent diffusion transformer

    Shuang Wu, Youtian Lin, Feihu Zhang, Yifei Zeng, Jingxi Xu, Philip Torr, Xun Cao, and Yao Yao. Direct3d: Scalable image-to-3d generation via 3d latent diffusion transformer.arXiv preprint arXiv:2405.14832, 2024. 16

    work page arXiv 2024

  43. [43]

    Cat3d: Create any- thing in 3d with multi-view diffusion models,

    Ruiqi Gao, Aleksander Holynski, Philipp Henzler, Arthur Brussee, Ricardo Martin-Brualla, Pratul Srinivasan, Jonathan T. Barron, and Ben Poole. Cat3d: Create anything in 3d with multi-view diffusion models.arXiv preprint arXiv:2405.10314, 2024. 16

    work page arXiv 2024

  44. [44]

    Genassets: Generating in-the-wild 3d assets in latent space

    Ze Yang, Jingkang Wang, Haowei Zhang, Sivabalan Manivasagam, Yun Chen, and Raquel Urtasun. Genassets: Generating in-the-wild 3d assets in latent space. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition ( CVPR), 2025. 16

    2025

  45. [45]

    Qi, Mahyar Najibi, Boyang Deng, Leonidas Guibas, Yin Zhou, and Dragomir Anguelov

    Bokui Shen, Xinchen Yan, Charles R. Qi, Mahyar Najibi, Boyang Deng, Leonidas Guibas, Yin Zhou, and Dragomir Anguelov. Gina-3d: Learning to generate implicit neural assets in the wild.arXiv preprint arXiv:2304.02163, 2023. 16

    work page arXiv 2023

  46. [46]

    Dreamcar: Leveraging car-specific prior for in-the-wild 3d car reconstruction.IEEE Robotics and Automation Letters, 10(2):1840–1847, 2024

    Xiaobiao Du, Haiyang Sun, Ming Lu, Tianqing Zhu, and Xin Yu. Dreamcar: Leveraging car-specific prior for in-the-wild 3d car reconstruction.IEEE Robotics and Automation Letters, 10(2):1840–1847, 2024. 16

    2024

  47. [47]

    Urbancad: Towards highly controllable and photorealistic 3d vehicles for urban scene simulation

    Yichong Lu, Yichi Cai, Shangzhan Zhang, Hongyu Zhou, Haoji Hu, Huimin Yu, Andreas Geiger, and Yiyi Liao. Urbancad: Towards highly controllable and photorealistic 3d vehicles for urban scene simulation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 27519–27530, 2025. 16

    2025

  48. [48]

    Rgm: Reconstructing high-fidelity 3d car assets with relightable 3d-gs generative model from a single image.arXiv preprint arXiv:2410.08181, 2024

    Xiaoxue Chen, Jv Zheng, Hao Huang, Haoran Xu, Weihao Gu, Kangliang Chen, Huan-ang Gao, Hao Zhao, Guyue Zhou, Yaqin Zhang, et al. Rgm: Reconstructing high-fidelity 3d car assets with relightable 3d-gs generative model from a single image.arXiv preprint arXiv:2410.08181, 2024. 16

    work page arXiv 2024

  49. [49]

    3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics ( SIGGRAPH), 2023

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk"uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics ( SIGGRAPH), 2023. 16

    2023

  50. [50]

    Neural scene graphs for dynamic scenes

    Julian Ost, Fahim Mannan, Nils Thuerey, Julian Knodt, and Felix Heide. Neural scene graphs for dynamic scenes. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition ( CVPR), 2021. 16

    2021

  51. [51]

    Omnire: Omni urban scene reconstruction

    Ziyu Chen, Jiawei Yang, Jiahui Huang, Riccardo de Lutio, Janick Martinez Esturo, Boris Ivanovic, Or Litany, Zan Gojcic, Sanja Fidler, Marco Pavone, Li Song, and Yue Wang. Omnire: Omni urban scene reconstruction. InThe Thirteenth International Conference on Learning Representations, 2025. URL https://openreview.net/forum?id=11xgiMEI5o. 16 21 Asset Harveste...

    2025

  52. [52]

    Carla documentation: Nvidia nurec

    CARLA Team. Carla documentation: Nvidia nurec. Website, 2026. URL https://carla.readthedocs.io/en/ latest/nvidia_nurec/. Accessed 2026-03-09. 17

    2026

  53. [53]

    Alpasim: A modular, lightweight, and data-driven research simulator for autonomous driving

    NVIDIA, Yulong Cao, Riccardo de Lutio, Sanja Fidler, Guillermo Garcia Cobo, Zan Gojcic, Maximilian Igl, Boris Ivanovic, Peter Karkus, Janick Martinez Esturo, Marco Pavone, Aaron Smith, Ellie Tanimura, Michal Tyszkiewicz, Michael Watson, Qi Wu, and Le Zhang. Alpasim: A modular, lightweight, and data-driven research simulator for autonomous driving. Softwar...

    2025