Pith. sign in

REVIEW 2 major objections 4 minor 99 references

Novel views with absolute global camera control can be synthesized from ordinary unposed photos by conditioning an image editor on target poses in object coordinates plus a text definition of the front face.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-12 07:33 UTC pith:7KH4AWHI

load-bearing objection Clean absolute-pose NVS from unposed inputs with real fidelity gains; the SOTA ranking vs relative methods is partly an evaluation artifact, but the capability itself is still useful. the 2 major comments →

arxiv 2607.02712 v1 pith:7KH4AWHI submitted 2026-07-02 cs.CV

Global Pose Control for Generative View Synthesis in Normalized Object Coordinate Space

Zhibing Li , Amogh Gupta , Behnoosh Parsa , Dan Casas This is my paper
classification cs.CV
keywords novel view synthesiscamera pose controlnormalized object coordinate spaceimage editingdiffusion modelsPlücker ray mapscanonical orientationunposed images
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Existing generative novel-view methods either control the camera only relative to the input photo or produce sparse global views, so users cannot freely request a standard front, side or top view of an object. This paper shows that the problem can be solved by working entirely in a normalized object coordinate space: the model receives one or a few unposed images, the absolute target camera pose expressed in that space, and a short sentence that defines which side of the object is the front. By casting the task as image editing rather than pure generation, the system inherits the high visual fidelity of modern editing models while the camera is injected as dedicated ray-map tokens under a regional attention mask that binds each pose to its own output view. The result is consistently oriented, high-quality novel views across many categories, without ever needing the poses of the input images. A new cleaned collection of 22 thousand orientation-aligned objects with matching front-face captions supplies the necessary training signal.

Core claim

Formulating novel view synthesis as instruction-based image editing and conditioning solely on absolute target camera poses in Normalized Object Coordinate Space (encoded as Plücker ray maps) together with a natural-language definition of the object’s front face yields accurate global viewpoint control from single or few unposed images, outperforming both image-to-3D pipelines and relative-pose generative baselines in fidelity and orientation consistency.

What carries the argument

In-context multi-modal camera conditioning with regional attention: target poses become compact Plücker ray-map tokens packed along the frame axis of a multi-modal diffusion transformer; a regional attention mask forces each camera token to attend only to its matched target view while text anchors the canonical front of the object.

Load-bearing premise

That a filtered set of synthetic 3D objects plus automatically written front-face captions is clean and diverse enough for the learned mapping to transfer to real photographs and multi-object scenes without extra pose supervision.

What would settle it

On a held-out collection of real photographs of everyday objects whose true front faces are unambiguous, command a set of absolute NOCS poses and measure whether the generated views match ground-truth orientations and appearance more accurately than strong relative-pose baselines given estimated input poses; systematic flips or large metric drops would falsify the claim.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Downstream systems that need a shared canonical frame (robotics, simulation, content pipelines) can request standard views without ever estimating input camera poses.
  • Users can redefine the object’s front face at test time simply by changing the text prompt, reorienting the entire coordinate system without retraining.
  • High-fidelity 2D image quality is retained while multi-view geometric consistency and global orientation awareness are added.
  • The newly curated 22k-object dataset with front-face captions becomes a reusable resource for any method that requires consistent object orientation.

Where Pith is reading between the lines

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

  • The same regional-attention ray-map tokens could be inserted into video diffusion backbones to supply absolute camera control without relying on temporal continuity priors that conflict with unordered viewpoints.
  • Residual failures on multi-object scenes and complex geometry imply that an explicit object-centric isolation or multi-NOCS stage is a natural next architectural step.
  • Because orientation is supplied by text, category-specific fine-tuning may become unnecessary for many practical capture-to-view pipelines.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper proposes a generative novel-view synthesis method that conditions solely on absolute target camera poses in a Normalized Object Coordinate Space (NOCS) plus a natural-language definition of the object’s front face, using one or few unposed input images. NVS is cast as instruction-based image editing on a LoRA-adapted Qwen-Image-Edit MMDiT; Plücker ray-map tokens are packed along the frame axis with RoPE interpolation and bound to their target views by a regional attention mask. A 22 K-object orientation-aligned Objaverse subset (plus Objaverse-OA and ABO) is curated with VLM front-face captions. Quantitative results on GSO and Toys4k claim state-of-the-art fidelity and orientation accuracy over both image-to-3D and relative-pose generative NVS baselines.

Significance. If the absolute-NOCS formulation and regional-attention binding hold under fair evaluation, the work supplies a practical route to global, text-customizable viewpoint control without requiring input poses or intermediate 3D meshes. The curated 22 K-object set with explicit front-face text definitions is a reusable resource. Ablations (Table 3, Figs. 7–9) cleanly isolate the contributions of ray-map tokens, regional attention, and front-view anchoring, and the method inherits the high visual fidelity of a modern image-editing backbone—advantages that remain valuable even after the evaluation protocol is equalized.

major comments (2)
  1. [§4.3, Table 2, Appendix D.2] Table 2 / Fig. 6 and the abstract claim of SOTA over generative NVS methods rest on an asymmetric protocol: EscherNet and SEVA receive Orient-Anything-V2 pose estimates while the proposed method needs none. Appendix D.2 shows that, once SEVA is given ground-truth poses on Toys4k, its multi-view PSNR exceeds the proposed method (20.17 / 21.35 vs. 18.93 / 19.31 for 2 / 3 views). The main-text ranking and the “error accumulation” narrative are therefore not fully supported; either the main tables must be re-ranked under GT poses or the claim must be restated as “pose-free absolute control under estimated-pose conditions.”
  2. [Abstract, §4, Limitations, Fig. 9(b)] The Limitations section and Fig. 9(b) acknowledge degradation on multi-object scenes and complex geometric compositions, yet the abstract and §4 still assert “robust o across diverse categories.” Because the training corpus is a filtered single-object Objaverse subset, the transfer claim needs either quantitative multi-object / real-world metrics or a clear scope restriction; otherwise the strongest claim over-reaches the evidence.
minor comments (4)
  1. [Limitations] Clarify whether camera intrinsics are assumed identical across views (stated only in Limitations) and how the method would be extended if they vary.
  2. [§3.3] The power-weighted and truncated-exponential view-sampling distributions (Eqs. 2–3) are given without sensitivity analysis; a short ablation on k and λ would strengthen reproducibility.
  3. [Fig. 1, Appendix C.2] Figure 1 caption and the “Orientation Matters” column would benefit from an explicit note that the 3D baselines are rendered after best-of-four front-direction selection (as described in Appendix C.2).
  4. [Abstract / front matter] Typographical inconsistencies appear in the arXiv header (“togeneratingsparseglobalviews”, missing spaces) and should be cleaned for the camera-ready version.

Circularity Check

0 steps flagged

No circularity: empirical systems paper whose claims rest on held-out metrics, not self-referential equations or load-bearing self-citations.

full rationale

The paper is a standard CV systems contribution: it defines an architecture (Plücker raymap tokens + regional attention inside a LoRA-adapted MMDiT, Eq. 1 and §3.2), trains with ordinary flow-matching loss (Eq. 4) on a curated orientation-aligned dataset (§3.4), and reports PSNR/SSIM/LPIPS/CLIP against held-out GSO and Toys4k renders (Tables 1–2). No equation reduces a claimed “prediction” to a fitted constant by construction; no uniqueness theorem is imported from overlapping authors; no ansatz is smuggled via self-citation. Dataset filtering and VLM captions are external preprocessing, not part of the evaluation loop. Evaluation asymmetry versus relative-pose baselines (Appendix D.2) is a fairness issue, not circularity. The derivation chain is therefore self-contained and non-circular.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 2 invented entities

As an empirical deep-learning systems paper the load-bearing premises are standard domain assumptions of diffusion models plus a handful of engineering choices (LoRA rank, ray-map resolution, power-law view sampling). No new physical entities are postulated; the ‘invented’ pieces are architectural modules whose utility is measured by ablation.

free parameters (4)
  • LoRA rank = 64
    Set to 64; controls capacity of the adaptation and is not derived from first principles.
  • learning rate = 1e-4
    AdamW learning rate 1e-4 chosen by the authors; affects convergence of the reported metrics.
  • ray-map resolution = 256×256
    Fixed at 256×256 independent of image resolution; an engineering trade-off between token length and spatial fidelity.
  • view-sampling exponents (k=1.5, λ=ln2) = k=1.5, λ=ln2
    Power-law and exponential distributions that bias the training distribution toward harder multi-view configurations; chosen by hand.
axioms (4)
  • domain assumption A pretrained instruction-based image-editing MMDiT (Qwen-Image-Edit) already possesses sufficient appearance priors that LoRA fine-tuning can add geometric control without destroying fidelity.
    Stated in §3.1–3.2; the entire method inherits quality from this base model.
  • domain assumption Plücker ray maps at reduced resolution, after RoPE interpolation, are a sufficient absolute camera representation for the diffusion transformer.
    §3.2; alternative encodings are not exhaustively compared.
  • domain assumption Textual front-face definitions generated by a VLM are accurate enough to disambiguate NOCS for both training and zero-shot categories.
    §3.4 and ablation in §4.4; residual VLM errors would propagate into orientation mistakes.
  • ad hoc to paper Camera intrinsics are identical across all views and can be held fixed.
    Explicitly listed as a limitation; the model is never trained with varying focal length.
invented entities (2)
  • Regional attention mask for camera–view binding no independent evidence
    purpose: Prevents cross-view leakage when multiple target ray maps and images share the same spatial grid under 3-D RoPE.
    Introduced in §3.2; ablation shows large metric drop without it. No independent theoretical derivation; utility is empirical.
  • Canonical front-view anchoring during joint multi-view generation no independent evidence
    purpose: Forces the model to resolve the object’s front before synthesizing other poses, acting as an implicit chain-of-thought.
    §3.3; ablation on swords shows orientation flips without it.

pith-pipeline@v1.1.0-grok45 · 25088 in / 2942 out tokens · 36261 ms · 2026-07-12T07:33:08.921721+00:00 · methodology

0 comments
read the original abstract

Novel View Synthesis (NVS) enables the generation of unseen views of a scene from a single or multiple images, allowing users to freely explore an object from any viewpoint. Despite the recent impressive qualitative improvements of generative models for this task, existing methods struggle to provide global and intuitive control of target viewpoints because they either use input-relative camera poses or are limited to generating sparse global views. This lack of global pose control severely limits the number of downstream tasks potentially enabled by NVS. To address this limitation, we propose a novel approach for precise camera control in a customizable Normalized Object Coordinate Space (NOCS), requiring single or few unposed images. Our method operates solely on the absolute camera pose of the target view in NOCS, eliminating the need for a relative world frame or camera poses of the input images. Unlike previous methods that treat NVS as a standalone generation task, we formulate it as an image editing problem and build upon state-of-the-art editing models to leverage their superior generalization capability. Camera information is injected as dedicated camera tokens via an in-context multi-modal conditioning strategy. To alleviate the inherent ambiguity of NOCS, we incorporate text descriptions that explicitly define the object's canonical coordinate frame, which also enhances generalization to unseen object categories. Furthermore, we curate a high-quality dataset with consistently aligned orientations and corresponding NOCS text definitions. Extensive experiments demonstrate that our method robustly generates novel views with accurate and consistent orientations from arbitrary unposed images across diverse categories, achieving state-of-the-art image quality and fidelity.

Figures

Figures reproduced from arXiv: 2607.02712 by Amogh Gupta, Behnoosh Parsa, Dan Casas, Zhibing Li.

Figure 1
Figure 1. Figure 1: Existing NVS methods fall short in providing global viewpoint control and high￾quality appearance: image-to-3D methods (4th column) lack orientation awareness; orientation-aware image-to-3D (5th column) produce coarse textures; and 2D-based NVS (6th column) only provide input-relative camera control. In contrast, our method achieves both global viewpoint control and high image fidelity, with the front face… view at source ↗
Figure 2
Figure 2. Figure 2: Given input images captured from arbitrary viewpoints (left) and target camera poses (top-right), our model generates high-quality oriented novel views. Notice how the generated images maintain correct global orientation across different object families (e.g., the front of the object aligned with the red axis). Furthermore, we leverage the native text-conditioning capabilities of the un￾derlying editing mo… view at source ↗
Figure 3
Figure 3. Figure 3: Overview. Left: Target camera poses are encoded as Plücker ray map tokens and fed into a LoRA-adapted MMDiT alongside image tokens. A text prompt defines the object’s canonical orientation in NOCS. Top-right: Target images and ray map tokens are packed along the frame axis. The position embeddings of ray map tokens are interpolated to spatially align with the corresponding image tokens. Bottom-right: Regio… view at source ↗
Figure 4
Figure 4. Figure 4: Common failure modes in Objaverse-OA. (a) objects with imprecise orientations (e.g., tilted or rotated off-axis), (b) ob￾jects with no recognizable or am￾biguous front views, and (c) ob￾jects with over-simplified textures. Our approach requires canonically aligned, high-quality 3D objects with front-face an￾notations at training time. Unfortunately, no existing dataset fully satisfies these require￾ments: … view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison with image-to-3D methods. [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative comparison with generative NVS methods. [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Ablation study on camera conditioning. Qwen-Image-Edit cannot ro￾bustly produce multi-view layouts without fine-tuning. LoRA enables layout genera￾tion, but text-only camera descriptions lack precision. Plücker raymap tokens provide explicit geometric conditioning, yet full attention causes cross-view leakage. Regional attention restricts each raymap to its corresponding view, eliminating leakage and achie… view at source ↗
Figure 8
Figure 8. Figure 8: Ablation study on front view definition. [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Ablation study on canonical front view anchoring and Limitations. [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
Figure 1
Figure 1. Figure 1: Ours (no pose input) vs. SEVA (w/ GT input pose). alignment — though our method delivers visibly better overall generation quality ( [PITH_FULL_IMAGE:figures/full_fig_p026_1.png] view at source ↗

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

99 extracted references · 18 linked inside Pith

  1. [1]

    Anthropic News (2025),https://www.anthropic

    Anthropic: Claude opus 4.5. Anthropic News (2025),https://www.anthropic. com/news/claude-sonnet-4-5

    2025

  2. [2]

    In: International Conference on Learning Representations (ICLR) (2025),https: //openreview.net/forum?id=0n4bS0R5MM

    Bahmani, S., Skorokhodov, I., Siarohin, A., Menapace, W., Qian, G., Vasilkovsky, M., Lee, H.Y., Wang, C., Zou, J., Tagliasacchi, A., Lindell, D.B., Tulyakov, S.: VD3d: Taming large video diffusion transformers for 3d camera control. In: International Conference on Learning Representations (ICLR) (2025),https: //openreview.net/forum?id=0n4bS0R5MM

    2025

  3. [3]

    arXiv preprint arXiv:2503.11647 (2025)

    Bai,J., Xia, M., Fu, X.,Wang, X.,Mu, L., Cao, J.,Liu, Z., Hu, H.,Bai, X., Wan, P., et al.: Recammaster: Camera-controlled generative rendering from a single video. arXiv preprint arXiv:2503.11647 (2025)

    Pith/arXiv arXiv 2025

  4. [4]

    Bai, S., Chen, K., Liu, X., Wang, J., Ge, W., Song, S., Dang, K., Wang, P., Wang, S., Tang, J., Zhong, H., Zhu, Y., Yang, M., Li, Z., Wan, J., Wang, P., Ding, W., Fu, Z., Xu, Y., Ye, J., Zhang, X., Xie, T., Cheng, Z., Zhang, H., Yang, Z., Xu, H., Lin, J.: Qwen2.5-vl technical report (2025),https://arxiv.org/abs/2502.13923

    Pith/arXiv arXiv 2025

  5. [5]

    In: arXiv (2021)

    Chan, E.R., Lin, C.Z., Chan, M.A., Nagano, K., Pan, B., Mello, S.D., Gallo, O., Guibas, L., Tremblay, J., Khamis, S., Karras, T., Wetzstein, G.: Efficient geometry- aware 3D generative adversarial networks. In: arXiv (2021)

    2021

  6. [6]

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

    Charatan, D., Li, S., Tagliasacchi, A., Sitzmann, V.: pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 20912– 20922 (2024)

    2024

  7. [7]

    In: European Conference on Computer Vision (ECCV) (2020) 16 Z

    Chen, X., Dong, Z., Song, J., Geiger, A., Hilliges, O.: Category level object pose estimation via neural analysis-by-synthesis. In: European Conference on Computer Vision (ECCV) (2020) 16 Z. Li et al

    2020

  8. [8]

    In: European Conference on Computer Vision

    Chen, Y., Xu, H., Zheng, C., Zhuang, B., Pollefeys, M., Geiger, A., Cham, T.J., Cai, J.: Mvsplat: Efficient 3d gaussian splatting from sparse multi-view images. In: European Conference on Computer Vision. Springer (2024)

    2024

  9. [9]

    CVPR (2022)

    Collins, J., Goel, S., Deng, K., Luthra, A., Xu, L., Gundogdu, E., Zhang, X., Yago Vicente, T.F., Dideriksen, T., Arora, H., Guillaumin, M., Malik, J.: Abo: Dataset and benchmarks for real-world 3d object understanding. CVPR (2022)

    2022

  10. [10]

    Conference on Neural Information Processing Systems (NeurIPS)36, 35799–35813 (2023)

    Deitke, M., Liu, R., Wallingford, M., Ngo, H., Michel, O., Kusupati, A., Fan, A., Laforte, C., Voleti, V.S., Gadre, S.Y., VanderBilt, E., Kembhavi, A., Vondrick, C., Gkioxari, G., Ehsani, K., Schmidt, L., Farhadi, A.: Objaverse-XL: A Uni- verse of 10M+ 3D Objects. Conference on Neural Information Processing Systems (NeurIPS)36, 35799–35813 (2023)

    2023

  11. [11]

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

    Deitke, M., Schwenk, D., Salvador, J., Weihs, L., Michel, O., VanderBilt, E., Schmidt, L., Ehsani, K., Kembhavi, A., Farhadi, A.: Objaverse: A Universe of An- notated 3D Objects. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 13142–13153 (2022)

    2022

  12. [12]

    International Conference on Robotics and Au- tomation (ICRA) pp

    Downs, L., Francis, A., Koenig, N., Kinman, B., Hickman, R.M., Reymann, K., McHugh, T.B., Vanhoucke, V.: Google Scanned Objects: A High-Quality Dataset of 3D Scanned Household Items. International Conference on Robotics and Au- tomation (ICRA) pp. 2553–2560 (2022)

    2022

  13. [13]

    In: Forty-first international conference on machine learning (2024)

    Esser, P., Kulal, S., Blattmann, A., Entezari, R., Müller, J., Saini, H., Levi, Y., Lorenz, D., Sauer, A., Boesel, F., et al.: Scaling rectified flow transformers for high-resolution image synthesis. In: Forty-first international conference on machine learning (2024)

    2024

  14. [14]

    Feng, J., Li, X., Lin, J., Liu, J., Liu, G., Lou, W., Ma, S., Shi, G., Wang, Q., Wang, J., Xu, Z., Yi, X., Yu, Z., Zhang, J., Zhu, Y., Chen, R., Chi, J., Du, Z., Han, L., Huang, L., Jiang, K., Li, Y., Luo, G., Wang, S., Wu, Q., Yang, F., Zhang, J., Zhang, X.: Seed3d 1.0: From images to high-fidelity simulation-ready 3d assets (2025),https://arxiv.org/abs/...

    arXiv 2025

  15. [15]

    Advances in Neural Information Processing Systems (2024)

    Gao*, R., Holynski*, A., Henzler, P., Brussee, A., Martin-Brualla, R., Srinivasan, P.P., Barron, J.T., Poole*, B.: Cat3d: Create anything in 3d with multi-view dif- fusion models. Advances in Neural Information Processing Systems (2024)

    2024

  16. [16]

    In: Advances in neural information processing systems

    Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., Bengio, Y.: Generative adversarial networks. In: Advances in neural information processing systems. vol. 27, pp. 2672–2680 (2014)

    2014

  17. [17]

    In: Pro- ceedings of the 23rd annual conference on Computer graphics and interactive tech- niques

    Gortler, S.J., Grzeszczuk, R., Szeliski, R., Cohen, M.F.: The lumigraph. In: Pro- ceedings of the 23rd annual conference on Computer graphics and interactive tech- niques. pp. 43–54 (1996)

    1996

  18. [18]

    He, H., Xu, Y., Guo, Y., Wetzstein, G., Dai, B., Li, H., Yang, C.: Cameractrl: Enabling camera control for text-to-video generation (2024)

    2024

  19. [19]

    In: Advances in Neural Information Processing Systems

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. In: Advances in Neural Information Processing Systems. vol. 33, pp. 6840–6851 (2020)

    2020

  20. [20]

    In: International Conference on Learning Representations (2024)

    Hong, Y., Zhang, K., Gu, J., Bi, S., Yang, Y., Forsyth, D., et al.: Lrm: Large recon- struction model for single image to 3d. In: International Conference on Learning Representations (2024)

    2024

  21. [21]

    In: International Con- ference on Learning Representations (2022),https://openreview.net/forum?id= nZeVKeeFYf9

    Hu, E.J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., Chen, W.: LoRA: Low-rank adaptation of large language models. In: International Con- ference on Learning Representations (2022),https://openreview.net/forum?id= nZeVKeeFYf9

    2022

  22. [22]

    net/forum?id=qSCjC9PFhs Generative View Synthesis in Normalized Object Coordinate Space 17

    Huang, B., Duan, H., Zhao, Y., Zhao, Z., Ma, Y., Gao, S.: CUPID: Pose-grounded generative 3d reconstruction from a single image (2025),https://openreview. net/forum?id=qSCjC9PFhs Generative View Synthesis in Normalized Object Coordinate Space 17

    2025

  23. [23]

    arXiv preprint arxiv:2410.23775 (2024)

    Huang,L.,Wang,W.,Wu,Z.F.,Shi,Y.,Dou,H.,Liang,C.,Feng,Y.,Liu,Y.,Zhou, J.: In-Context LoRA for Diffusion Transformers. arXiv preprint arxiv:2410.23775 (2024)

    Pith/arXiv arXiv 2024

  24. [24]

    Hunyuan3D, T., :, Zhang, B., Guo, C., Guo, D., Liu, H., Yan, H., Shi, H., Yu, J., Xu, J., Huang, J., Li, K., Wang, L., Linus, Wang, P., Lin, Q., Tang, R., Yang, X., Li, Y., Guan, Y., Zhao, Y., Yang, Y., Lai, Z., Liang, Z., Zhao, Z.: Hy3d-bench: Generation of 3d assets (2026),https://arxiv.org/abs/2602.03907

    arXiv 2026

  25. [25]

    In: International Conference on Learning Representations (2025)

    Jin, H., Jiang, H., Tan, H., Zhang, K., Bi, S., Zhang, T., Luan, F., Snavely, N., Xu, Z.: Lvsm: A large view synthesis model with minimal 3d inductive bias. In: International Conference on Learning Representations (2025)

    2025

  26. [26]

    ACM Transactions on Graphics (ToG)42(4), 1–14 (2023)

    Kerbl, B., Kopanas, G., Leimkühler, T., Drettakis, G.: 3d gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics (ToG)42(4), 1–14 (2023)

    2023

  27. [27]

    arXiv preprint arXiv:2402.03908 (2024)

    Kong,X.,Liu,S.,Lyu,X.,Taher,M.,Qi,X.,Davison,A.J.:Eschernet:Agenerative model for scalable view synthesis. arXiv preprint arXiv:2402.03908 (2024)

    Pith/arXiv arXiv 2024

  28. [28]

    In: Proceedings of the 23rd annual conference on Computer graphics and interactive techniques

    Levoy, M., Hanrahan, P.: Light field rendering. In: Proceedings of the 23rd annual conference on Computer graphics and interactive techniques. pp. 31–42 (1996)

    1996

  29. [29]

    arXiv preprint arXiv:2507.10496 (2025)

    Li, R., Yi, B., Liu, J., Gao, H., Ma, Y., Kanazawa, A.: Cameras as relative posi- tional encoding. arXiv preprint arXiv:2507.10496 (2025)

    arXiv 2025

  30. [30]

    arXiv preprint arXiv:2508.20470 (2025)

    Li, X., Du, G., Zhang, R., Jin, L., Jia, Q., Lu, L., Guo, Z., Zhao, Y., Liu, H., Wang, T., Li, C., Gong, X., Li, R., Fan, B.: Droplet3D: Commonsense Priors from Videos Facilitate 3D Generation. arXiv preprint arXiv:2508.20470 (2025)

    Pith/arXiv arXiv 2025

  31. [31]

    arXiv preprint arXiv:2505.23253 (2025)

    Liang, Y., Luo, K., Chen, X., Chen, R., Yan, H., Li, W., Liu, J., Tan, P.: Uni- tex: Universal high fidelity generative texturing for 3d shapes. arXiv preprint arXiv:2505.23253 (2025)

    Pith/arXiv arXiv 2025

  32. [32]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024),https://arxiv.org/pdf/2311.07885

    Liu, M., Shi, R., Chen, L., Zhang, Z., Xu, C., Wei, X., Chen, H., Zeng, C., Gu, J., Su, H.: One-2-3-45++: Fast Single Image to 3D Objects with Consistent Multi- View Generation and 3D Diffusion. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024),https://arxiv.org/pdf/2311.07885. pdf

    Pith/arXiv arXiv 2024

  33. [33]

    Conference on Neural Information Processing Systems (NeurIPS)36(2024)

    Liu, M., Xu, C., Jin, H., Chen, L., Varma T, M., Xu, Z., Su, H.: One-2-3-45: Any single image to 3d mesh in 45 seconds without per-shape optimization. Conference on Neural Information Processing Systems (NeurIPS)36(2024)

    2024

  34. [34]

    In: International Conference on Computer Vision (ICCV) (2023)

    Liu, R., Wu, R., Hoorick, B.V., Tokmakov, P., Zakharov, S., Vondrick, C.: Zero-1- to-3: Zero-shot One Image to 3D Object. In: International Conference on Computer Vision (ICCV) (2023)

    2023

  35. [35]

    In: Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision

    Liu, R., Wu, R., Van Hoorick, B., Tokmakov, P., Zakharov, S., Vondrick, C.: Zero- 1-to-3: Zero-shot one image to 3d object. In: Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision. pp. 9298–9309 (2023)

    2023

  36. [36]

    Liu, S., Ng, K.W., Jang, W., Guo, J., Han, J., Liu, H., Douratsos, Y., Pérez, J.C., Zhou, Z., Phung, C., et al.: Scaling sequence-to-sequence generative neural rendering.In:InternationalConferenceonLearningRepresentations(ICLR)(2026)

    2026

  37. [37]

    In: International Conference on Learning Representations (ICLR) (2024),https: //openreview.net/forum?id=MN3yH2ovHb

    Liu, Y., Lin, C., Zeng, Z., Long, X., Liu, L., Komura, T., Wang, W.: Sync- Dreamer: Generating Multiview-consistent Images from a Single-view Image. In: International Conference on Learning Representations (ICLR) (2024),https: //openreview.net/forum?id=MN3yH2ovHb

    2024

  38. [38]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024) 18 Z

    Long, X., Guo, Y.C., Lin, C., Liu, Y., Dou, Z., Liu, L., Ma, Y., Zhang, S.H., Habermann, M., Theobalt, C., et al.: Wonder3D: Single Image to 3D using Cross- Domain Diffusion. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024) 18 Z. Li et al

    2024

  39. [39]

    Lu, Y., Tian, Y., Jiang, Z., Zhao, Y., Yang, Y., Ouyang, H., Hu, H., Yu, H., Shen, Y., Liao, Y.: Orientation matters: Making 3d generative models orientation-aligned (2025),https://arxiv.org/abs/2506.08640

    arXiv 2025

  40. [40]

    International Conference on Machine Learning, 2024 (2024)

    Melas-Kyriazi, L., Laina, I., Rupprecht, C., Neverova, N., Vedaldi, A., Gafni, O., Kokkinos, F.: Im-3d: Iterative multiview diffusion and reconstruction for high- quality 3d generation. International Conference on Machine Learning, 2024 (2024)

    2024

  41. [41]

    In: Eu- ropean Conference on Computer Vision

    Mildenhall, B., Srinivasan, P.P., Tancik, M., Barron, J.T., Ramamoorthi, R., Ng, R.: Nerf: Representing scenes as neural radiance fields for view synthesis. In: Eu- ropean Conference on Computer Vision. pp. 405–421. Springer (2020)

    2020

  42. [42]

    arXiv preprint arXiv:2310.10375 (2023)

    Miyato, T., Jaeger, B., Welling, M., Geiger, A.: Gta: A geometry-aware attention mechanism for multi-view transformers. arXiv preprint arXiv:2310.10375 (2023)

    Pith/arXiv arXiv 2023

  43. [43]

    ACM Transactions on Graphics (ToG)41(4), 1– 15 (2022)

    Müller,T.,Evans,A.,Schied,C.,Keller,A.:Instantneuralgraphicsprimitiveswith a multiresolution hash encoding. ACM Transactions on Graphics (ToG)41(4), 1– 15 (2022)

    2022

  44. [44]

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

    Peng, S., Liu, Y., Huang, Q.X., Bao, H., Zhou, X.: Pvnet: Pixel-wise voting net- work for 6dof pose estimation. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 4556–4565 (2018)

    2018

  45. [45]

    In: International Conference on Learning Representations (2023)

    Poole, B., Jain, A., Barron, J.T., Mildenhall, B.: Dreamfusion: Text-to-3d using 2d diffusion. In: International Conference on Learning Representations (2023)

    2023

  46. [46]

    In: International conference on machine learning

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

    2021

  47. [47]

    Advances in neural information processing systems36, 53728–53741 (2023)

    Rafailov, R., Sharma, A., Mitchell, E., Manning, C.D., Ermon, S., Finn, C.: Direct preference optimization: Your language model is secretly a reward model. Advances in neural information processing systems36, 53728–53741 (2023)

    2023

  48. [48]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2025)

    Ren,X.,Shen,T.,Huang,J.,Ling,H.,Lu,Y.,Nimier-David,M.,Müller,T.,Keller, A., Fidler, S., Gao, J.: GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2025)

    2025

  49. [49]

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

    Sajjadi, M.S.M., Meyer, H., Pot, E., Bergmann, U., Greff, K., Radwan, N., Vora, S., Lučić, M., Duckworth, D., Dosovitskiy, A., Uszkoreit, J., Funkhouser, T., Tagliasacchi, A.: Scene representation transformer: Geometry-free novel view syn- thesis through set-latent scene representations. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pa...

    2022

  50. [50]

    arXiv preprint arXiv:2402.03300 (2024)

    Shao, Z., Wang, P., Zhu, Q., Xu, R., Song, J., Bi, X., Zhang, H., Zhang, M., Li, Y., Wu, Y., et al.: Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300 (2024)

    Pith/arXiv arXiv 2024

  51. [51]

    Shi, L., Shi, Y., Xu, X., Liu, T., Xi, J., Chen, C.: NVSPolicy: Adaptive Novel- View Synthesis for Generalizable Language-Conditioned Policy Learning (2025), https://arxiv.org/abs/2505.10359

    Pith/arXiv arXiv 2025

  52. [52]

    Shi, R., Chen, H., Zhang, Z., Liu, M., Xu, C., Wei, X., Chen, L., Zeng, C., Su, H.: Zero123++: a Single Image to Consistent Multi-view Diffusion Base Model (2023)

    2023

  53. [53]

    In: International Conference on Learning Representations (ICLR) (2024),https://openreview.net/forum?id=FUgrjq2pbB

    Shi, Y., Wang, P., Ye, J., Mai, L., Li, K., Yang, X.: MVDream: Multi-view Diffu- sion for 3D Generation. In: International Conference on Learning Representations (ICLR) (2024),https://openreview.net/forum?id=FUgrjq2pbB

    2024

  54. [54]

    Song, Y., Ermon, S.: Generative modeling by estimating gradients of the data distribution.In:AdvancesinNeuralInformationProcessingSystems.vol.32(2019)

    2019

  55. [55]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2021) Generative View Synthesis in Normalized Object Coordinate Space 19

    Stojanov, S., Thai, A., Rehg, J.M.: Using Shape to Categorize: Low-Shot Learning with an Explicit Shape Bias. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2021) Generative View Synthesis in Normalized Object Coordinate Space 19

    2021

  56. [56]

    In: IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR)

    Sun, J., Wang, Z., Zhang, S., He, X.H., Zhao, H., Zhang, G., Zhou, X.: OnePose: One-Shot Object Pose Estimation without CAD Models. In: IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR). pp. 6815–6824 (2022)

    2022

  57. [57]

    In: International Conference on Computer Vision (ICCV)

    Szymanowicz, S., Rupprecht, C., Vedaldi, A.: Viewset Diffusion: (0-)Image- Conditioned 3D Generative Models from 2D Data. In: International Conference on Computer Vision (ICCV). pp. 8829–8839 (2023)

    2023

  58. [58]

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

    Szymanowicz, S., Rupprecht, C., Vedaldi, A.: Splatter image: Ultra-fast single-view 3d reconstruction. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 20892–20901 (2024)

    2024

  59. [59]

    Team, T.H.: Hunyuan3D 2.1: From Images to High-Fidelity 3D Assets with Production-Ready PBR Material (2025)

    2025

  60. [60]

    In: IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR) (2023),https://arxiv.org/abs/ 2303.17598

    Tseng, H.Y., Li, Q., Kim, C., Alsisan, S., Huang, J.B., Kopf, J.: Consistent view synthesis with pose-guided diffusion models. In: IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR) (2023),https://arxiv.org/abs/ 2303.17598

    Pith/arXiv arXiv 2023

  61. [61]

    Van Hoorick, B., Wu, R., Ozguroglu, E., Sargent, K., Liu, R., Tokmakov, P., Dave, A., Zheng, C., Vondrick, C.: Generative camera dolly: Extreme monocular dy- namicnovelviewsynthesis.In:EuropeanConferenceonComputerVision.Springer (2024)

    2024

  62. [62]

    In: European Conference on Computer Vision (ECCV) (2024)

    Voleti, V., Yao, C.H., Boss, M., Letts, A., Pankratz, D., Tochilkin, D., Laforte, C., Rombach, R., Jampani, V.: SV3D: Novel Multi-view Synthesis and 3D Generation from a Single Image using Latent Video Diffusion. In: European Conference on Computer Vision (ECCV) (2024)

    2024

  63. [63]

    arXiv preprint arXiv:2503.20314 (2025)

    Wan, T., Wang, A., Ai, B., Wen, B., Mao, C., Xie, C.W., Chen, D., Yu, F., Zhao, H., Yang, J., Zeng, J., Wang, J., Zhang, J., Zhou, J., Wang, J., Chen, J., Zhu, K., Zhao, K., Yan, K., Huang, L., Feng, M., Zhang, N., Li, P., Wu, P., Chu, R., Feng, R., Zhang, S., Sun, S., Fang, T., Wang, T., Gui, T., Weng, T., Shen, T., Lin, W., Wang, W., Wang, W., Zhou, W.,...

    Pith/arXiv arXiv 2025

  64. [64]

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

    Wang, H., Du, X., Li, J., Yeh, R.A., Shakhnarovich, G.: Score jacobian chaining: Lifting pretrained 2d diffusion models for 3d generation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 12619– 12629 (2023)

    2023

  65. [65]

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

    Wang, H., Sridhar, S., Huang, J., Valentin, J., Song, S., Guibas, L.J.: Normalized object coordinate space for category-level 6d object pose and size estimation. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 2642–2651 (2019)

    2019

  66. [66]

    arXiv preprint arXiv:2312.02201 (2023)

    Wang, P., Shi, Y.: Imagedream: Image-prompt multi-view diffusion for 3d genera- tion. arXiv preprint arXiv:2312.02201 (2023)

    Pith/arXiv arXiv 2023

  67. [67]

    In: Inter- national Conference on Machine Learning (ICML) (2025),https://openreview

    Wang, Z., Zhang, Z., Pang, T., Du, C., Zhao, H., Zhao, Z.: Orient Anything: Learn- ing Robust Object Orientation Estimation from Rendering 3D Models. In: Inter- national Conference on Machine Learning (ICML) (2025),https://openreview. net/forum?id=x4yTgv2WkJ

    2025

  68. [68]

    Li et al

    Wang, Z., Zhang, Z., Xu, J., Wang, J., Pang, T., Du, C., Zhao, H., Zhao, Z.: Orient anythingv2:Unifyingorientationandrotationunderstanding.In:TheThirty-ninth Annual Conference on Neural Information Processing Systems (2026) 20 Z. Li et al

    2026

  69. [69]

    In: ACM SIGGRAPH 2024 Conference Papers (2023)

    Wang, Z., Yuan, Z., Wang, X., Li, Y., Chen, T., Xia, M., Luo, P., Shan, Y.: Mo- tionctrl: A unified and flexible motion controller for video generation. In: ACM SIGGRAPH 2024 Conference Papers (2023)

    2024

  70. [70]

    In: International Conference on Learning Representations (ICLR) (2023),https://openreview.net/forum?id= HtoA0oT30jC

    Watson, D., Chan, W., Brualla, R.M., Ho, J., Tagliasacchi, A., Norouzi, M.: Novel View Synthesis with Diffusion Models. In: International Conference on Learning Representations (ICLR) (2023),https://openreview.net/forum?id= HtoA0oT30jC

    2023

  71. [71]

    Advances in neural information processing systems35, 24824–24837 (2022)

    Wei, J., Wang, X., Schuurmans, D., Bosma, M., Xia, F., Chi, E., Le, Q.V., Zhou, D., et al.: Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems35, 24824–24837 (2022)

    2022

  72. [72]

    In: IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR)

    Wen, B., Yang, W., Kautz, J., Birchfield, S.T.: FoundationPose: Unified 6D Pose Estimation and Tracking of Novel Objects. In: IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR). pp. 17868–17879 (2023)

    2023

  73. [73]

    arXiv preprint arXiv:2509.20328 (2025)

    Wiedemer, T., Li, Y., Vicol, P., Gu, S.S., Matarese, N., Swersky, K., Kim, B., Jaini, P., Geirhos, R.: Video models are zero-shot learners and reasoners. arXiv preprint arXiv:2509.20328 (2025)

    Pith/arXiv arXiv 2025

  74. [74]

    Wu, C., Li, J., Zhou, J., Lin, J., Gao, K., Yan, K., ming Yin, S., Bai, S., Xu, X., Chen, Y., Chen, Y., Tang, Z., Zhang, Z., Wang, Z., Yang, A., Yu, B., Cheng, C., Liu, D., Li, D., Zhang, H., Meng, H., Wei, H., Ni, J., Chen, K., Cao, K., Peng, L., Qu, L., Wu, M., Wang, P., Yu, S., Wen, T., Feng, W., Xu, X., Wang, Y., Zhang, Y., Zhu, Y., Wu, Y., Cai, Y., L...

    Pith/arXiv arXiv 2025

  75. [75]

    In: Advances in Neural Information Processing Systems (NeurIPS)

    Wu, J., Zhang, C., Xue, T., Freeman, W.T., Tenenbaum, J.B.: Learning a prob- abilistic latent space of object shapes via 3d generative-adversarial modeling. In: Advances in Neural Information Processing Systems (NeurIPS). vol. 29, pp. 82–90 (2016)

    2016

  76. [76]

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

    Wu, R., Mildenhall, B., Henzler, P., Park, K., Gao, R., Watson, D., Srinivasan, P.P., Verbin, D., Barron, J.T., Poole, B., et al.: Reconfusion: 3d reconstruction with diffusion priors. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 21382–21392 (2024)

    2024

  77. [77]

    Tech report (2025)

    Xiang, J., Chen, X., Xu, S., Wang, R., Lv, Z., Deng, Y., Zhu, H., Dong, Y., Zhao, H., Yuan, N.J., Yang, J.: Native and compact structured latents for 3d generation. Tech report (2025)

    2025

  78. [78]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2025),https://arxiv.org/abs/2412.01506

    Xiang, J., Lv, Z., Xu, S., Deng, Y., Wang, R., Zhang, B., Chen, D., Tong, X., Yang, J.: Structured 3D Latents for Scalable and Versatile 3D Generation. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2025),https://arxiv.org/abs/2412.01506

    Pith/arXiv arXiv 2025

  79. [79]

    In: European Conference on Computer Vision (ECCV) (2024)

    Xu, Y., Shi, Z., Yifan, W., Chen, H., Yang, C., Peng, S., Shen, Y., Wetzstein, G.: GRM: Large Gaussian Reconstruction Model for Efficient 3D Reconstruction and Generation. In: European Conference on Computer Vision (ECCV) (2024)

    2024

  80. [80]

    arXiv preprint arXiv:2408.06072 (2024)

    Yang, Z., Teng, J., Zheng, W., Ding, M., Huang, S., Xu, J., Yang, Y., Hong, W., Zhang, X., Feng, G., et al.: Cogvideox: Text-to-video diffusion models with an expert transformer. arXiv preprint arXiv:2408.06072 (2024)

    Pith/arXiv arXiv 2024

Showing first 80 references.