pith. sign in

arxiv: 2606.13652 · v1 · pith:EEZP2E7Gnew · submitted 2026-06-11 · 💻 cs.CV · cs.GR

World Tracing: Generative Pixel-Aligned Geometry Beyond the Visible

Hao Zhang , Mohamed El Banani , Jen-Hao Cheng , Paul Zhang , Yi Hua , Ben Mildenhall , Christoph Lassner , Narendra Ahuja
show 1 more author
Gengshan Yang
This is my paper

Pith reviewed 2026-06-27 06:54 UTC · model grok-4.3

classification 💻 cs.CV cs.GR
keywords pixel-aligned geometrygenerative 3D reconstructionoccluded geometrydiffusion transformerimage-to-3Ddepth estimation3D scene completion
0
0 comments X

The pith

World Tracing predicts ordered stacks of camera-space 3D points per pixel to reconstruct visible surfaces and generate occluded geometry while remaining pixel-aligned.

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

The paper introduces a representation that outputs multiple 3D points for every input pixel, ordered from front to back in camera space. The first point matches the visible surface while later points complete hidden surfaces that depth estimators normally omit. This is realized by training a diffusion transformer to denoise the full stack under a mixed noise schedule that balances reconstruction accuracy with generative completion. The approach targets the trade-off where depth methods stay faithful to pixels but stop at the visible surface, and full image-to-3D generators produce complete shapes that drift from the input image. If successful, the method supplies geometry that can be edited or synthesized while preserving exact 2D-to-3D correspondence.

Core claim

World Tracing is a generative pixel-aligned geometry representation that predicts an ordered stack of camera-space 3D points for each input pixel, where the first layer represents the visible surface and subsequent layers represent front-to-back intersections with occluded surfaces. The representation is instantiated as a world-tracing diffusion transformer (WT-DiT) that treats multiple geometry layers as separate denoising tokens coupled through factorized and global attention, trained with pixel-space flow matching and a mixed noise schedule.

What carries the argument

The ordered stack of camera-space 3D points per pixel, with the diffusion transformer treating geometry layers as separate denoising tokens coupled by factorized and global attention.

If this is right

  • Outperforms both depth predictors and image-to-3D generators on visible-surface reconstruction and complete geometry generation across object, scene, and dynamic benchmarks.
  • Preserves 2D-to-3D correspondence, supporting text-driven 3D scene editing.
  • Enables geometry-conditioned novel-view video synthesis.
  • Allows training-free integration with textured-mesh generators.

Where Pith is reading between the lines

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

  • The per-pixel stack format could be extended to video sequences by adding a temporal dimension to the attention mechanism, allowing consistent completion across frames.
  • Because each point remains tied to an input pixel, the representation may reduce drift in single-view SLAM systems when occluded surfaces are needed for loop closure.
  • The separation of visible and occluded layers suggests a natural way to condition downstream tasks such as semantic segmentation on only the visible layer while still having access to full scene geometry.

Load-bearing premise

A single diffusion transformer with factorized and global attention, trained under a mixed noise schedule, can simultaneously achieve accurate visible-surface reconstruction and plausible occluded-geometry generation without layer collapse or misalignment artifacts.

What would settle it

On a held-out multi-view benchmark, the first-layer points deviate from ground-truth visible depth by more than a few pixels on average, or the later layers fail to produce intersections consistent with hidden surfaces observed from other viewpoints.

Figures

Figures reproduced from arXiv: 2606.13652 by Ben Mildenhall, Christoph Lassner, Gengshan Yang, Hao Zhang, Jen-Hao Cheng, Mohamed El Banani, Narendra Ahuja, Paul Zhang, Yi Hua.

Figure 1
Figure 1. Figure 1: World Tracing. A pixel-aligned layered geometry representation that faithfully generates complete objects, scenes, and dynamic content from single images and monocular videos. Colored points are predicted visible surfaces; gray points are predicted surfaces hidden from inputs. We visualize depth in magma colormap. This pixel-aligned representation enables several downstream applications: training-free pose… view at source ↗
Figure 2
Figure 2. Figure 2: WT-DiT architecture. A frozen MoGe encoder provides pixel-aligned image features, while noisy multilayer XYZ is patchified into geometry tokens. Pixel-aligned fusion combines image and geometry tokens before DiT decoder blocks with layer-wise, ray-wise, and global self-attention, plus temporal attention for WT-D. A linear patch projection maps each decoder token to the XYZ of its 14×14 patch, followed by u… view at source ↗
Figure 3
Figure 3. Figure 3: Pixel-aligned geometry as a unified 3D interface. WT generates complete multilayer geometry in the input camera frame while preserving pixel correspondences. This representation enables pose-aware structure for training-free textured-mesh pipelines such as TRELLIS-style decoders, and serves as geometry memory for novel-view video synthesis. matching loss, depth filling, frozen image encoder, and structured… view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative comparison on out-of-distribution inputs. All inputs are deliberately drawn from outside our training distributions to probe generalization. Top: object examples comparing WT-O against LaRI-O and TRELLIS.2; inputs are real-world DAVIS [45] video frames and generated images, neither of which appear in our object training corpus. Bottom: scene examples comparing WT-S against LaRI-S and MoGe-2 on … view at source ↗
Figure 5
Figure 5. Figure 5: Multilayer depth stack produced by WT-S. Each row shows one input (left) followed by the six predicted depth layers in turbo colormap. Top two rows are held-out 3D-FRONT frames; the remaining seven are out-of-distribution generated indoor rooms. As ℓ increases, occluded geometry behind near surfaces is filled in (e.g. floor and walls behind furniture, room interiors behind doorways and chandeliers) while L… view at source ↗
Figure 6
Figure 6. Figure 6: Training timestep distributions. Analytic densities of the standard logit-normal schedule, the plateaued logit-normal variant, and their equal-weight 50/50 mixture used in phase (ii) of the layer-aware curriculum. C Data Pipeline Details This appendix expands Sec. 3.4 with the full data-pipeline specification. Rendered inputs and source datasets. We render RGBA images with randomized lighting, view￾points,… view at source ↗
read the original abstract

Image-to-3D methods often trade off faithfulness and completeness: depth estimators are anchored to input pixels but stop at the visible surface, while image-to-3D models generate complete shapes that are often misaligned with the input. We introduce World Tracing, a generative pixel-aligned geometry representation that predicts 3D points aligned with observed pixels while completing geometry beyond the visible surface. For each input pixel, World Tracing predicts an ordered stack of camera-space 3D points, where the first layer represents the visible surface and subsequent layers represent front-to-back intersections with occluded surfaces. We instantiate this representation with a world-tracing diffusion transformer, WT-DiT, which treats multiple geometry layers as separate denoising tokens coupled through factorized and global attention. WT-DiT is trained with pixel-space flow matching and a mixed noise schedule that balances visible-surface reconstruction with occluded-geometry generation. World Tracing achieves strong performance on visible-surface reconstruction and complete geometry generation across object, scene, and dynamic benchmarks, outperforming both depth predictors and image-to-3D generators. It also preserves 2D-to-3D correspondence, enabling text-driven 3D scene editing, geometry-conditioned novel-view video synthesis, and training-free integration with textured-mesh generators.

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 / 2 minor

Summary. The paper introduces World Tracing, a generative pixel-aligned geometry representation that, for each input pixel, predicts an ordered stack of camera-space 3D points where layer 1 is the visible surface and later layers are front-to-back occluded intersections. The representation is realized via a world-tracing diffusion transformer (WT-DiT) that treats geometry layers as separate denoising tokens coupled by factorized and global attention; the model is trained with pixel-space flow matching under a mixed noise schedule intended to balance visible-surface reconstruction and occluded-geometry generation. The manuscript claims that this approach achieves strong performance on visible-surface reconstruction and complete geometry generation across object, scene, and dynamic benchmarks, outperforming both depth predictors and image-to-3D generators while preserving 2D-to-3D correspondence for downstream tasks such as text-driven editing and geometry-conditioned novel-view synthesis.

Significance. If the empirical results and ordering guarantees hold, the work provides a concrete mechanism for reconciling pixel faithfulness with geometric completeness, which has been a persistent tension in image-to-3D literature. The ordered multi-layer tokenization and mixed-schedule training constitute a technically interesting architectural choice that could be reusable beyond the specific benchmarks. The preservation of explicit 2D-to-3D correspondence is a practical strength that directly enables the listed editing and synthesis applications.

major comments (2)
  1. [Abstract] Abstract: the central claim of outperforming baselines on multiple benchmarks is stated without any quantitative metrics, tables, error bars, ablation results, or dataset specifications. Because the soundness of the performance claim is load-bearing for the paper's contribution, the absence of these numbers prevents evaluation of whether the ordered-stack representation actually delivers the advertised gains over depth estimators and image-to-3D generators.
  2. [Method] Method (WT-DiT training description): the mixed noise schedule is presented as the sole mechanism that balances visible-surface accuracy with occluded-layer generation and enforces ordering. No additional loss term, ordering regularizer, or explicit supervision for layer correspondence is mentioned; if the schedule does not sufficiently differentiate noise levels across layers or if factorized attention fails to propagate pixel alignment, later layers can collapse or misalign, directly violating the ordered-stack representation that underpins all downstream claims.
minor comments (2)
  1. The phrase 'pixel-space flow matching' is used without a reference or brief definition; a short citation to the relevant flow-matching formulation would improve readability.
  2. [Abstract] The abstract lists three application scenarios (text-driven editing, novel-view video synthesis, training-free mesh integration) but does not indicate whether these are demonstrated with qualitative figures or quantitative metrics in the main text.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. The comments correctly identify areas where additional clarity and quantitative support would strengthen the manuscript. We address each major comment below and outline the revisions we will make.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim of outperforming baselines on multiple benchmarks is stated without any quantitative metrics, tables, error bars, ablation results, or dataset specifications. Because the soundness of the performance claim is load-bearing for the paper's contribution, the absence of these numbers prevents evaluation of whether the ordered-stack representation actually delivers the advertised gains over depth estimators and image-to-3D generators.

    Authors: We agree that the abstract would be strengthened by including key quantitative results. The main body contains the supporting tables, error bars, ablations, and dataset specifications. We will revise the abstract to incorporate representative metrics (with references to the relevant tables) so that the performance claims are more directly evaluable from the abstract alone. revision: yes

  2. Referee: [Method] Method (WT-DiT training description): the mixed noise schedule is presented as the sole mechanism that balances visible-surface accuracy with occluded-layer generation and enforces ordering. No additional loss term, ordering regularizer, or explicit supervision for layer correspondence is mentioned; if the schedule does not sufficiently differentiate noise levels across layers or if factorized attention fails to propagate pixel alignment, later layers can collapse or misalign, directly violating the ordered-stack representation that underpins all downstream claims.

    Authors: The mixed noise schedule is the primary mechanism described for balancing the two objectives while the factorized and global attention layers are intended to maintain pixel alignment and inter-layer consistency. No explicit ordering regularizer or additional loss term is present in the current manuscript. We will expand Section 3 to provide a more detailed derivation of how the schedule differentiates noise levels across layers, include an ablation on ordering stability when the schedule is ablated, and explicitly discuss the role of attention in preventing layer collapse. This will make the training procedure more transparent and address the potential failure modes raised. revision: yes

Circularity Check

0 steps flagged

No circularity: new representation and training procedure are self-contained

full rationale

The paper defines World Tracing as an ordered per-pixel stack of camera-space points and instantiates it via a new WT-DiT architecture trained end-to-end with pixel-space flow matching under a mixed noise schedule on external datasets. No equations, loss terms, or architectural choices are shown to reduce by construction to quantities fitted from the authors' prior work; the ordering, alignment, and completion objectives are enforced by the training procedure itself rather than presupposed. The abstract and described method contain no self-citation load-bearing steps that justify the central claims, satisfying the criteria for a self-contained derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no equations, training details, or architectural specifications sufficient to identify free parameters, axioms, or invented entities.

pith-pipeline@v0.9.1-grok · 5776 in / 1134 out tokens · 25228 ms · 2026-06-27T06:54:57.117599+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

92 extracted references · 7 linked inside Pith

  1. [1]

    ARKitScenes: A diverse real-world dataset for 3D indoor scene understanding using mobile RGB-D data

    Gilad Baruch, Zhuoyuan Chen, Afshin Dehghan, Tal Dimry, Yuri Feigin, Peter Fu, Thomas Gebauer, Brandon Joffe, Daniel Kurz, Arik Schwartz, and Elad Shulman. ARKitScenes: A diverse real-world dataset for 3D indoor scene understanding using mobile RGB-D data. InNeurIPS Datasets and Benchmarks Track, 2021

    2021

  2. [2]

    FLUX.2: Analyzing and enhancing the latent space of FLUX – representation comparison.https://bfl.ai/techblog/representation-comparison/, 2025

    Black Forest Labs. FLUX.2: Analyzing and enhancing the latent space of FLUX – representation comparison.https://bfl.ai/techblog/representation-comparison/, 2025

    2025

  3. [3]

    FreeOrbit4D: Training-free arbitrary camera redirection for monocular videos via geometry-complete 4D reconstruction.arXiv preprint arXiv:2601.18993, 2026

    Wei Cao, Hao Zhang, Fengrui Tian, Yulun Wu, Yingying Li, Shenlong Wang, Ning Yu, and Yaoyao Liu. FreeOrbit4D: Training-free arbitrary camera redirection for monocular videos via geometry-complete 4D reconstruction.arXiv preprint arXiv:2601.18993, 2026

    Pith/arXiv arXiv 2026

  4. [4]

    ReconViaGen: Towards accurate multi-view 3D object reconstruction via generation.arXiv preprint arXiv:2510.23306, 2025

    Jiahao Chang, Chongjie Ye, Yushuang Wu, Yuantao Chen, Yidan Zhang, Zhongjin Luo, Chenghong Li, Yihao Zhi, and Xiaoguang Han. ReconViaGen: Towards accurate multi-view 3D object reconstruction via generation.arXiv preprint arXiv:2510.23306, 2025

    arXiv 2025

  5. [5]

    Chang, Manolis Savva, Maciej Halber, Thomas Funkhouser, and Matthias Nießner

    Angela Dai, Angel X. Chang, Manolis Savva, Maciej Halber, Thomas Funkhouser, and Matthias Nießner. ScanNet: Richly-annotated 3D reconstructions of indoor scenes. InCVPR, 2017

    2017

  6. [6]

    Objaverse-XL: A universe of 10m+ 3d objects

    Matt Deitke, Ruoshi Liu, Matthew Wallingford, Huong Ngo, Oscar Michel, Aditya Kusupati, Alan Fan, Christian Laforte, Vikram Voleti, Samir Yitzhak Gadre, et al. Objaverse-XL: A universe of 10m+ 3d objects. InNeurIPS, 2023

    2023

  7. [7]

    Objaverse: Auniverseofannotated 3d objects

    Matt Deitke, Dustin Schwenk, Jordi Salvador, Luca Weihs, Oscar Michel, Eli VanderBilt, Ludwig Schmidt,KianaEhsani,AniruddhaKembhavi,andAliFarhadi. Objaverse: Auniverseofannotated 3d objects. InCVPR, 2023

    2023

  8. [8]

    McHugh, and Vincent Vanhoucke

    Laura Downs, Anthony Francis, Nathan Koenig, Brandon Kinman, Ryan Hickman, Krista Rey- mann, Thomas B. McHugh, and Vincent Vanhoucke. Google Scanned Objects: A high-quality dataset of 3d scanned household items. InICRA, 2022

    2022

  9. [9]

    Depth map prediction from a single image using a multi-scale deep network

    David Eigen, Christian Puhrsch, and Rob Fergus. Depth map prediction from a single image using a multi-scale deep network. InNeurIPS, 2014

    2014

  10. [10]

    Scaling rectified flow transformers for high-resolution image synthesis

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

    2024

  11. [11]

    Interactive order-independent transparency.White paper , nVIDIA, 2(6):7, 2001

    Cass Everitt. Interactive order-independent transparency.White paper , nVIDIA, 2(6):7, 2001

    2001

  12. [12]

    3D-FRONT: 3d furnished rooms with layouts and semantics

    Huan Fu, Bowen Cai, Lin Gao, Lingxiao Zhang, Jiaming Wang, Cao Li, Qixun Zeng, Chengyue Sun, Rongfei Jia, Binqiang Zhao, and Hao Zhang. 3D-FRONT: 3d furnished rooms with layouts and semantics. InICCV, 2021

    2021

  13. [13]

    3D-FUTURE: 3D furniture shape with texture.IJCV, 2021

    Huan Fu, Rongfei Jia, Lin Gao, Mingming Gong, Binqiang Zhao, Steve Maybank, and Dacheng Tao. 3D-FUTURE: 3D furniture shape with texture.IJCV, 2021

    2021

  14. [14]

    GeoWizard: Unleashing the diffusion priors for 3d geometry estimation from a single image.ECCV, 2024

    Xiao Fu, Wei Yin, Mu Hu, Kaixuan Wang, Yuexin Ma, Ping Tan, Shaojie Shen, Dahua Lin, and Xiaoxiao Long. GeoWizard: Unleashing the diffusion priors for 3d geometry estimation from a single image.ECCV, 2024

    2024

  15. [15]

    Lotus: Diffusion-based visual foundation model for high-quality dense prediction.ICLR, 2025

    Jing He, Haodong Li, Wei Yin, Yixun Liang, Leheng Li, Kaiqiang Zhou, Hongbo Zhang, Bingbing Liu, and Ying-Cong Chen. Lotus: Diffusion-based visual foundation model for high-quality dense prediction.ICLR, 2025

    2025

  16. [16]

    LRM: Large reconstruction model for single image to 3d.ICLR, 2024

    Yicong Hong, Kai Zhang, Jiuxiang Gu, Sai Bi, Yang Zhou, Difan Liu, Feng Liu, Kalyan Sunkavalli, Trung Bui, and Hao Tan. LRM: Large reconstruction model for single image to 3d.ICLR, 2024. 13

    2024

  17. [17]

    X-ray: A sequential 3d representation for generation, 2024

    Tao Hu, Wenhang Ge, Yuyang Zhao, and Gim Hee Lee. X-ray: A sequential 3d representation for generation, 2024. URLhttps://arxiv.org/abs/2404.14329

    arXiv 2024

  18. [18]

    StructLDM: Structured latent diffusion for 3d human generation.ECCV, 2024

    Tao Hu, Fangzhou Hong, and Ziwei Liu. StructLDM: Structured latent diffusion for 3d human generation.ECCV, 2024

    2024

  19. [19]

    Consistent4D: Consistent 360° dynamic object generation from monocular video.ICLR, 2024

    Yanqin Jiang, Li Zhang, Jin Gao, Weiming Hu, and Yao Yao. Consistent4D: Consistent 360° dynamic object generation from monocular video.ICLR, 2024

    2024

  20. [20]

    VACE: All-in-one video creation and editing

    Zeyinzi Jiang, Zhen Han, Chaojie Mao, Jingfeng Zhang, Yulin Pan, and Yu Liu. VACE: All-in-one video creation and editing. InICCV, 2025

    2025

  21. [21]

    Peek-a- boo: Occlusion reasoning in indoor scenes with plane representations.CVPR, 2020

    Ziyu Jiang, Buyu Liu, Samuel Schulter, Zhangyang Wang, and Manmohan Chandraker. Peek-a- boo: Occlusion reasoning in indoor scenes with plane representations.CVPR, 2020

    2020

  22. [22]

    DualPM: Dual Posed-Canonical point maps for 3D shape and pose reconstruction

    Ben Kaye, Tomas Jakab, Shangzhe Wu, Christian Rupprecht, and Andrea Vedaldi. DualPM: Dual Posed-Canonical point maps for 3D shape and pose reconstruction. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 6425–6435, June 2025

    2025

  23. [23]

    Repurposing diffusion-based image generators for monocular depth estimation

    Bingxin Ke, Anton Obukhov, Shengyu Huang, Nando Metzger, Rodrigo Caye Daudt, and Konrad Schindler. Repurposing diffusion-based image generators for monocular depth estimation. In CVPR, 2024

    2024

  24. [24]

    Modular primitives for high-performance differentiable rendering.ACM Trans

    Samuli Laine, Janne Hellsten, Tero Karras, Yeongho Seol, Jaakko Lehtinen, and Timo Aila. Modular primitives for high-performance differentiable rendering.ACM Trans. Graph., 2020

    2020

  25. [25]

    Grounding image matching in 3d with MASt3R

    Vincent Leroy, Yohann Cabon, and Jerome Revaud. Grounding image matching in 3d with MASt3R. InECCV, 2024

    2024

  26. [26]

    LaRI: Layered ray intersec- tions for single-view 3d geometric reasoning.arXiv preprint arXiv:2504.18424, 2025

    Rui Li, Biao Zhang, Zhenyu Li, Federico Tombari, and Peter Wonka. LaRI: Layered ray intersec- tions for single-view 3d geometric reasoning.arXiv preprint arXiv:2504.18424, 2025

    Pith/arXiv arXiv 2025

  27. [27]

    Back to basics: Let denoising generative models denoise.arXiv preprint arXiv:2511.13720, 2025

    Tianhong Li and Kaiming He. Back to basics: Let denoising generative models denoise.arXiv preprint arXiv:2511.13720, 2025

    Pith/arXiv arXiv 2025

  28. [28]

    MegaDepth: Learning single-view depth prediction from Internet photos

    Zhengqi Li and Noah Snavely. MegaDepth: Learning single-view depth prediction from Internet photos. InCVPR, 2018

    2018

  29. [29]

    MegaSaM: Accurate, fast, and robust structure and motion from casual dynamic videos.CVPR, 2025

    Zhengqi Li, Richard Tucker, Forrester Cole, Qianqian Wang, Linyi Jin, Vickie Ye, Angjoo Kanazawa, Aleksander Holynski, and Noah Snavely. MegaSaM: Accurate, fast, and robust structure and motion from casual dynamic videos.CVPR, 2025

    2025

  30. [30]

    SS4D: Native 4d generative model via structured spacetime latents.ACM Trans

    Zhibing Li, Mengchen Zhang, Tong Wu, Jing Tan, Jiaqi Wang, and Dahua Lin. SS4D: Native 4d generative model via structured spacetime latents.ACM Trans. Graph., 2025

    2025

  31. [31]

    Wonderland: Navigating 3D scenes from a single image.arXiv preprint arXiv:2412.12091, 2024

    Hanwen Liang, Junli Cao, Vidit Goel, Guocheng Qian, Sergei Korolev, Demetri Terzopoulos, Konstantinos Plataniotis, Sergey Tulyakov, and Jian Ren. Wonderland: Navigating 3D scenes from a single image.arXiv preprint arXiv:2412.12091, 2024

    arXiv 2024

  32. [32]

    PAD3R: Pose-aware dynamic 3D reconstruction from casual videos.SIGGRAPH Asia, 2025

    Ting-Hsuan Liao, Haowen Liu, Yiran Xu, Songwei Ge, Gengshan Yang, and Jia-Bin Huang. PAD3R: Pose-aware dynamic 3D reconstruction from casual videos.SIGGRAPH Asia, 2025

    2025

  33. [33]

    Magic3D: High-resolution text-to-3d content creation

    Chen-Hsuan Lin, Jun Gao, Luming Tang, Towaki Takikawa, Xiaohui Zeng, Xun Huang, Karsten Kreis, Sanja Fidler, Ming-Yu Liu, and Tsung-Yi Lin. Magic3D: High-resolution text-to-3d content creation. InCVPR, 2023

    2023

  34. [34]

    Chen, Zhenyu Li, Guang Shi, Jiashi Feng, and Bingyi Kang

    Haotong Lin, Sili Chen, Jun Hao Liew, Donny Y. Chen, Zhenyu Li, Guang Shi, Jiashi Feng, and Bingyi Kang. Depth anything 3: Recovering the visual space from any views.arXiv preprint arXiv:2511.10647, 2025. 14

    Pith/arXiv arXiv 2025

  35. [35]

    Yaron Lipman, Ricky T. Q. Chen, Heli Ben-Hamu, Maximilian Nickel, and Matt Le. Flow matching for generative modeling. InICLR, 2023

    2023

  36. [36]

    One-2-3-45++: Fast single image to 3d objects with consistent multi-view generation and 3d diffusion.CVPR, 2024

    Minghua Liu, Ruoxi Shi, Linghao Chen, Zhuoyang Zhang, Chao Xu, Xinyue Wei, Hansheng Chen, Chong Zeng, Jiayuan Gu, and Hao Su. One-2-3-45++: Fast single image to 3d objects with consistent multi-view generation and 3d diffusion.CVPR, 2024

    2024

  37. [37]

    Zero-1-to-3: Zero-shot one image to 3d object

    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. InICCV, 2023

    2023

  38. [38]

    Flow straight and fast: Learning to generate and transfer data with rectified flow

    Xingchao Liu, Chengyue Gong, and Qiang Liu. Flow straight and fast: Learning to generate and transfer data with rectified flow. InICLR, 2023

    2023

  39. [39]

    Chen, Jiashi Feng, Yu-Wing Tai, Chi-Keung Tang, and Bingyi Kang

    Xinhang Liu, Yuxi Xiao, Donny Y. Chen, Jiashi Feng, Yu-Wing Tai, Chi-Keung Tang, and Bingyi Kang. Trace anything: Representing any video in 4d via trajectory fields.ICLR, 2026

    2026

  40. [40]

    SyncDreamer: Generating multiview-consistent images from a single-view image.ICLR, 2024

    Yuan Liu, Cheng Lin, Zijiao Zeng, Xiaoxiao Long, Lingjie Liu, Taku Komura, and Wenping Wang. SyncDreamer: Generating multiview-consistent images from a single-view image.ICLR, 2024

    2024

  41. [41]

    Wonder3D: Single image to 3d using cross-domain diffusion.CVPR, 2024

    Xiaoxiao Long, Yuan-Chen Guo, Cheng Lin, Yuan Liu, Zhiyang Dou, Lingjie Liu, Yuexin Ma, Song-Hai Zhang, Marc Habermann, Christian Theobalt, and Wenping Wang. Wonder3D: Single image to 3d using cross-domain diffusion.CVPR, 2024

    2024

  42. [42]

    Transparency and antialiasing algorithms implemented with the virtual pixel maps technique.IEEE Computer graphics and Applications, 9(4):43–55, 1989

    Abraham Mammen. Transparency and antialiasing algorithms implemented with the virtual pixel maps technique.IEEE Computer graphics and Applications, 9(4):43–55, 1989

    1989

  43. [43]

    PC2: Projection-conditioned point cloud diffusion for single-image 3d reconstruction.arXiv preprint arXiv:2302.10668, 2023

    Luke Melas-Kyriazi, Christian Rupprecht, and Andrea Vedaldi. PC2: Projection-conditioned point cloud diffusion for single-image 3d reconstruction.arXiv preprint arXiv:2302.10668, 2023

    arXiv 2023

  44. [44]

    Vo, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al

    Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy V. Vo, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. DINOv2: Learning robust visual features without supervision.TMLR, 2024

    2024

  45. [45]

    A benchmark dataset and evaluation methodology for video object segmentation

    Federico Perazzi, Jordi Pont-Tuset, Brian McWilliams, Luc Van Gool, Markus Gross, and Alexan- der Sorkine-Hornung. A benchmark dataset and evaluation methodology for video object segmentation. InCVPR, 2016

    2016

  46. [46]

    UniDepth: Universal monocular metric depth estimation

    Luigi Piccinelli, Yung-Hsu Yang, Christos Sakaridis, Mattia Segu, Siyuan Li, Luc Van Gool, and Fisher Yu. UniDepth: Universal monocular metric depth estimation. InCVPR, 2024

    2024

  47. [47]

    Barron, and Ben Mildenhall

    Ben Poole, Ajay Jain, Jonathan T. Barron, and Ben Mildenhall. DreamFusion: Text-to-3d using 2d diffusion. InICLR, 2023

    2023

  48. [48]

    Towards robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer.TPAMI, 2022

    René Ranftl, Katrin Lasinger, David Hafner, Konrad Schindler, and Vladlen Koltun. Towards robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer.TPAMI, 2022

    2022

  49. [49]

    Susskind

    Mike Roberts, Jason Ramapuram, Anurag Ranjan, Atulit Kumar, Miguel Angel Bautista, Nathan Paczan, Russ Webb, and Joshua M. Susskind. Hypersim: A photorealistic synthetic dataset for holistic indoor scene understanding. InICCV, 2021

    2021

  50. [50]

    Mitra, and Tom Monnier

    Remy Sabathier, David Novotny, Niloy J. Mitra, and Tom Monnier. ActionMesh: Animated 3d mesh generation with temporal 3d diffusion.CVPR, 2026

    2026

  51. [51]

    Pifu: Pixel-aligned implicit function for high-resolution clothed human digitization

    Shunsuke Saito, Zeng Huang, Ryota Natsume, Shigeo Morishima, Angjoo Kanazawa, and Hao Li. Pifu: Pixel-aligned implicit function for high-resolution clothed human digitization. In Proceedings of the IEEE/CVF international conference on computer vision, pages 2304–2314, 2019

    2019

  52. [52]

    Pifuhd: Multi-level pixel-aligned implicit function for high-resolution 3d human digitization, 2020

    Shunsuke Saito, Tomas Simon, Jason Saragih, and Hanbyul Joo. Pifuhd: Multi-level pixel-aligned implicit function for high-resolution 3d human digitization, 2020. URLhttps://arxiv.org/ abs/2004.00452. 15

    arXiv 2020

  53. [53]

    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.arXiv prepr...

    Pith/arXiv arXiv 2025

  54. [54]

    Schönberger, Silvano Galliani, Torsten Sattler, Konrad Schindler, Marc Pollefeys, and Andreas Geiger

    Thomas Schöps, Johannes L. Schönberger, Silvano Galliani, Torsten Sattler, Konrad Schindler, Marc Pollefeys, and Andreas Geiger. A multi-view stereo benchmark with high-resolution images and multi-camera videos. InCVPR, 2017

    2017

  55. [55]

    Using layered depth images for interactive rendering

    Jonathan Shade. Using layered depth images for interactive rendering. InSIGGRAPH (tutorial), 1998

    1998

  56. [56]

    Layered depth images

    Jonathan Shade, Steven Gortler, Li-wei He, and Richard Szeliski. Layered depth images. In SIGGRAPH, 1998

    1998

  57. [57]

    Zero123++: A single image to consistent multi-view diffusion base model.arXiv, 2023

    Ruoxi Shi, Hansheng Chen, Zhuoyang Zhang, Minghua Liu, Chao Xu, Xinyue Wei, Linghao Chen, Chong Zeng, and Hao Su. Zero123++: A single image to consistent multi-view diffusion base model.arXiv, 2023

    2023

  58. [58]

    3d photography using context- aware layered depth inpainting.CVPR, 2020

    Meng-Li Shih, Shih-Yang Su, Johannes Kopf, and Jia-Bin Huang. 3d photography using context- aware layered depth inpainting.CVPR, 2020

    2020

  59. [59]

    Indoor segmentation and support inference from RGBD images

    Nathan Silberman, Derek Hoiem, Pushmeet Kohli, and Rob Fergus. Indoor segmentation and support inference from RGBD images. InECCV, 2012

    2012

  60. [60]

    LDM3D:Latentdiffusion model for 3d.arXiv preprint arXiv:2305.10853, 2023

    Gabriela Ben Melech Stan, Diana Wofk, Scottie Fox, Alex Redden, Will Saxton, Jean Yu, Estelle Aflalo, Shao-YenTseng, FabioNonato, MatthiasMuller, andVasudevLal. LDM3D:Latentdiffusion model for 3d.arXiv preprint arXiv:2305.10853, 2023

    arXiv 2023

  61. [61]

    Stefan Stojanov, Anh Thai, and James M. Rehg. Using shape to categorize: Low-shot learning with an explicit shape bias. InCVPR, 2021

    2021

  62. [62]

    Scalabilityinperceptionforautonomous driving: Waymo Open dataset

    Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla, Aurelien Chouard, Vijaysai Patnaik, Paul Tsui, James Guo, Yin Zhou, Yuning Chai, Benjamin Caine, Vijay Vasudevan, Wei Han, Jiquan Ngiam, Hang Zhao, Aleksei Timofeev, Scott Ettinger, Maxim Krivokon, Amy Gao, Aditya Joshi, Yu Zhang, JonathonShlens, ZhifengChen,andDragomirAnguelov. Scalabilityinperceptionforauto...

    2020

  63. [63]

    DreamGaussian: Generative Gaussian splatting for efficient 3d content creation.arXiv, 2023

    Jiaxiang Tang, Jiawei Ren, Hang Zhou, Ziwei Liu, and Gang Zeng. DreamGaussian: Generative Gaussian splatting for efficient 3d content creation.arXiv, 2023

    2023

  64. [64]

    LGM: Large multi-view Gaussian model for high-resolution 3d content creation.arXiv, 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, 2024

    2024

  65. [65]

    Improving and generalizing flow-based generative models with minibatch optimal transport

    Alexander Tong, Kilian Fatras, Nikolay Malkin, Guillaume Huguet, Yanlei Zhang, Jarrid Rector- Brooks, Guy Wolf, and Yoshua Bengio. Improving and generalizing flow-based generative models with minibatch optimal transport. InTMLR, 2024

    2024

  66. [66]

    Going deeper with image transformers

    Hugo Touvron, Matthieu Cord, Alexandre Sablayrolles, Gabriel Synnaeve, and Hervé Jégou. Going deeper with image transformers. InICCV, 2021

    2021

  67. [67]

    Truebones motions animation studios

    Truebones Motions Animation Studios. Truebones motions animation studios. https:// truebones.com, 2024

    2024

  68. [68]

    VGGT: Visual geometry grounded transformer

    Jianyuan Wang, Minghao Chen, Nikita Karaev, Andrea Vedaldi, Christian Rupprecht, and David Novotny. VGGT: Visual geometry grounded transformer. InCVPR, 2025. 16

    2025

  69. [69]

    MoGe: Unlocking accurate monocular geometry estimation for open-domain images with optimal training supervision.CVPR, 2025

    Ruicheng Wang, Sicheng Xu, Cassie Dai, Jianfeng Xiang, Yu Deng, Xin Tong, and Jiaolong Yang. MoGe: Unlocking accurate monocular geometry estimation for open-domain images with optimal training supervision.CVPR, 2025

    2025

  70. [70]

    MoGe-2: Accurate monocular geometry with metric scale and sharp details.arXiv preprint arXiv:2507.02546, 2025

    Ruicheng Wang, Sicheng Xu, Yue Dong, Yu Deng, Jianfeng Xiang, Zelong Lv, Guangzhong Sun, Xin Tong, and Jiaolong Yang. MoGe-2: Accurate monocular geometry with metric scale and sharp details.arXiv preprint arXiv:2507.02546, 2025

    Pith/arXiv arXiv 2025

  71. [71]

    DUSt3R: Geometric 3d vision made easy

    Shuzhe Wang, Vincent Leroy, Yohann Cabon, Boris Chidlovskii, and Jerome Revaud. DUSt3R: Geometric 3d vision made easy. InCVPR, 2024

    2024

  72. [72]

    Yifan Wang, Jianjun Zhou, Haoyi Zhu, Wenzheng Chang, Yang Zhou, Zizun Li, Junyi Chen, Jiangmiao Pang, Chunhua Shen, and Tong He.π3: Permutation-equivariant visual geometry learning.arXiv preprint arXiv:2507.13347, 2025

    Pith/arXiv arXiv 2025

  73. [73]

    Argoverse 2: Next generation datasets for self-driving perception and forecasting

    Benjamin Wilson, William Qi, Tanmay Agarwal, John Lambert, Jagjeet Singh, Siddhesh Khandel- wal, Bowen Pan, Ratnesh Kumar, Andrew Hartnett, Jhony Kaesemodel Pontes, Deva Ramanan, Peter Carr, and James Hays. Argoverse 2: Next generation datasets for self-driving perception and forecasting. InNeurIPS Datasets and Benchmarks Track, 2021

    2021

  74. [74]

    AnimateAnyMesh: Afeed-forward4Dfoundation model for text-driven universal mesh animation.arXiv preprint arXiv:2506.09982, 2025

    ZijieWu,ChaohuiYu,FanWang,andXiangBai. AnimateAnyMesh: Afeed-forward4Dfoundation model for text-driven universal mesh animation.arXiv preprint arXiv:2506.09982, 2025

    arXiv 2025

  75. [75]

    Native and compact structured latents for 3D generation.Tech report, 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.Tech report, 2025

    2025

  76. [76]

    Structured 3D latents for scalable and versatile 3D generation

    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. CVPR, 2025

    2025

  77. [77]

    SpatialTrackerV2: 3d point tracking made easy.ICCV, 2025

    Yuxi Xiao, Jianyuan Wang, Nan Xue, Nikita Karaev, Yuri Makarov, Bingyi Kang, Xing Zhu, Hujun Bao, Yujun Shen, and Xiaowei Zhou. SpatialTrackerV2: 3d point tracking made easy.ICCV, 2025

    2025

  78. [78]

    Pixel-perfect visual geometry estimation.arXiv preprint arXiv:2601.05246, 2026

    Gangwei Xu, Haotong Lin, Hongcheng Luo, Haiyang Sun, Bing Wang, Guang Chen, Sida Peng, Hangjun Ye, and Xin Yang. Pixel-perfect visual geometry estimation.arXiv preprint arXiv:2601.05246, 2026

    arXiv 2026

  79. [79]

    InstantMesh: Efficient 3d mesh generation from a single image with sparse-view large reconstruction models

    Jiale Xu, Weihao Cheng, Yiming Gao, Xintao Wang, Shenghua Gao, and Ying Shan. InstantMesh: Efficient 3d mesh generation from a single image with sparse-view large reconstruction models. arXiv, 2024

    2024

  80. [80]

    LaS- Comp: Zero-shot 3D completion with latent-spatial consistency.arXiv preprint arXiv:2602.18735, 2026

    Weilong Yan, Haipeng Li, Hao Xu, Nianjin Ye, Yihao Ai, Shuaicheng Liu, and Jingyu Hu. LaS- Comp: Zero-shot 3D completion with latent-spatial consistency.arXiv preprint arXiv:2602.18735, 2026

    arXiv 2026

Showing first 80 references.