Pith. sign in

REVIEW 2 major objections 4 minor 83 references

MVGBench: Comprehensive Benchmark for Multi-view Generation Models

T0 review · 2 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read MVGBench argues that multi-view generators should be judged by whether their outputs fuse into one coherent 3D object, and proposes a ground-truth-free 3D self-consistency metric to do it.

desk verdict A genuinely useful ground-truth-free benchmark for multi-view generation whose 3D consistency metric is clever and mostly well validated, but the 'outperforms all' claim for ViFiGen rests on a metric whose behavior on inconsistent outputs is only indirectly tested. read the letter →

arxiv 2507.00006 v1 pith:YFBNTXE5 submitted 2025-06-11 cs.GR cs.LGeess.IV

Xianghui Xie , Chuhang Zou , Meher Gitika Karumuri , Jan Eric Lenssen , Gerard Pons-Moll This is my paper
classification cs.GRcs.LGeess.IV
keywords multi-viewgeneration3DconsistencybenchmarkGaussianSplattingimage-to-3Dgenerativeevaluationvision-languagemodelViFiGen
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

MVGBench argues that multi-view image generation models should be judged by whether their outputs cohere into a single 3D object, not by how closely they match one ground-truth view. Its central proposal is a ground-truth-free 3D self-consistency metric: the generated views are split into two subsets, each subset is used to fit a 3D Gaussian Splatting, and the disagreement between the two reconstructions is treated as the measure of inconsistency. The benchmark adds image-quality and semantic metrics, evaluates twelve existing models on four curated real and synthetic datasets, and finds a persistent trade-off between 3D consistency and image quality as well as a large synthetic-to-real gap. Combining the design choices it identifies as best, the paper introduces ViFiGen, a video-based generator reported to outperform all evaluated models on 3D consistency.

What carries the argument

Split-then-reconstruct self-consistency comparison. Generated multi-views are divided into two subsets, two independent 3D Gaussian Splatting reconstructions (a point-based radiance-field representation of anisotropic Gaussians) are optimized, and their disagreement is computed as geometry errors (Chamfer distance on points resampled from the Gaussians, plus rendered depth error) and texture errors (PSNR, SSIM, and LPIPS between renderings from shared test views). Because all methods are compared through the same test views, with ICP alignment when no ground-truth scale exists, methods trained with different cameras, focal lengths, and elevation conventions can be ranked on the same scale.

What would settle it

Take ground-truth multi-view images, inject a controlled amount of geometric or texture inconsistency into one subset of views, and verify that the metric rises monotonically with the injected amount across view counts and elevations; any plateau or reversal would show that fitting artifacts, not generation inconsistency, drive the score.

Watch

Extended reading notes

Core claim

Fitting two 3D Gaussian Splattings to disjoint subsets of a generator's output views and measuring their geometric and texture disagreement gives a valid, fair measure of 3D consistency for multi-view generators. The paper validates this metric against perfectly consistent ground-truth views, showing it is nearly invariant to the number of views and the camera rendering settings, and it can be applied to real images where no 3D ground truth exists. Classic pair-wise metrics can rank an inconsistent generator above a consistent one; the self-consistency metric reverses such rankings. On this measure, video-diffusion models balance consistency and image quality better than image-diffusion models, and the paper's ViFiGen, a video-based model with a fine-grained convolution input encoder and a stronger camera embedding, achieves the best 3D consistency among the twelve methods evaluated.

Load-bearing premise

The metric assumes that the disagreement between the two fitted 3D models comes from the generator's inconsistency, not from the reconstruction algorithm failing differently on the two view subsets.

Editorial extensions

If this is right

  • Multi-view generators can be ranked on real photographs, because the metric requires no 3D ground truth and no method-specific target views.
  • A model can look strong under pair-wise PSNR/SSIM comparison yet fail the self-consistency check, so the two rankings are not interchangeable.
  • Video-diffusion backbones have a measurable advantage in the consistency-quality trade-off, giving future work a concrete starting point.
  • Replacing a category-level image encoder with a fine-grained convolution encoder, and using ray- or position-based camera embeddings, is a transferable recipe for better 3D consistency.
  • High-frequency detail loss in the latent autoencoder degrades consistency even before generation, identifying the autoencoder as a bottleneck for fine structures.

Reading between the lines

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

  • The split-then-reconstruct protocol is a general recipe that could be carried over to text-to-3D or multi-view video generators with little change, since it never needs a reference 3D model.
  • If 3D Gaussian Splatting fitting is fragile for very few views or extreme elevations, the score could partly measure reconstruction stability rather than generation consistency; a learned few-view reconstructor as the fitting stage would separate the two.
  • The large synthetic-to-real gap points to training data diversity as the next major lever, independently of architecture choices.
  • At roughly 2.7 hours per 100 objects, the metric is cheap enough to serve as an automated regression test for new releases of multi-view generators.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. MVGBench proposes a benchmark for multi-view image generation (MVG) models that evaluates ten metrics across three dimensions: 3D geometric/texture consistency, image quality, and semantic consistency. Its central novelty is a ground-truth-free 3D self-consistency metric: generated multi-views are split into two disjoint subsets, two 3D Gaussian Splatting (3DGS) models are fitted independently, and the discrepancy between them (Chamfer distance, depth error, cPSNR/cSSIM/cLPIPS) is used as a measure of 3D consistency. The benchmark covers 12 MVG methods on four datasets, with best-setup, real-image generalization, and robustness evaluations. Using the benchmark's design-choice analysis, the authors propose ViFiGen (SV3D plus camera positional encoding plus a ConvNextV2 encoder) and claim it outperforms all evaluated MVGs on 3D consistency. The metric is validated on ground-truth views (Table 1), and the VLM-based quality/semantic metrics and oFID are validated against user studies.

Significance. If the 3D consistency metric is valid, MVGBench fills a real gap: existing evaluations compare generated views against method-specific ground truth, which is inappropriate for generative tasks. The paper's strengths include a genuinely GT-free consistency protocol, a broad systematic comparison of 12 state-of-the-art methods on four datasets, externally validated VLM metrics (Pearson 0.69-0.92), and a user study showing oFID aligns with human rankings better than dataset-level FID. The design-choice analysis (camera embedding, input encoder, feature interaction, training data amount) is informative and the code/model release will be valuable. However, the central metric's validity on inconsistent generated inputs is the load-bearing assumption, and it is not yet established; the benchmark's usefulness and the headline 'outperforms all' claim depend on resolving this concern. The paper is a solid candidate for publication after substantial additional validation.

major comments (2)
  1. [Section 4.3 / Section 5 / Table 3] The design choices for ViFiGen were selected using the MVGBench analysis, and the same benchmark is then used to claim that ViFiGen outperforms all evaluated methods. This creates a mild self-referential loop: the benchmark is used both to discover best practices and to evaluate the method built from them. The benchmark is partly externally grounded (Table 1, VLM user studies), so this is not disqualifying, but the 'outperforms all' claim currently has no independent check. Please provide an external validation of this claim, such as a human study on 3D consistency or an evaluation on a held-out dataset not involved in the design-choice analysis; alternatively, explicitly frame the claim as 'in-benchmark' rather than universal.
  2. [Section 6.1 / Equation (5)] The normalization for CD, depth, and cLPIPS in Eq. (5) uses max_e_mvg, the maximum error over the evaluated methods, making the normalized robustness scores dependent on the composition of the evaluated method set. If the set of methods changes, the normalized values and possibly the method rankings can change. The paper should state this relativity explicitly and report raw scores or use a fixed reference (e.g., ground-truth-based upper bound) as the normalization anchor.
minor comments (4)
  1. [Section 4.1 / Table 1] The relative standard deviation for depth is 0.082 (8.2%), while the text states 'the deviation is less than 8% of the average score across all variants'; please reconcile the number and the claim.
  2. [Section 8] The sentence 'our method cannot evaluate methods that generate very few views (<10)' refers to the benchmark, not to the proposed ViFiGen model; please rephrase to avoid ambiguity.
  3. [References] Several references contain stray page numbers or malformed entries (e.g., references [2], [8], [9], and [54] end with extra page-like numbers); please clean the reference list.
  4. [Section 4.1 / Figure 5] The claim that VLM metrics strongly correlate with human perception is supported numerically only in Section 5's summary; please report the per-metric Pearson coefficients in Section 4.1 as well.

Circularity Check

1 steps flagged · score 4.0 of 10

Benchmark metrics are externally grounded, but ViFiGen is selected with MVGBench and then declared best on the same MVGBench ranking, a partial self-referential loop.

  1. fitted input called prediction [Abstract; Sec. 4.3 'MVG design choices', Tab. 3 vs Tab. 2]
    "Using the discovered best practices, we propose ViFiGen, a method that outperforms all evaluated MVGs on 3D consistency."

    In Sec. 4.3 the design choices for ViFiGen are chosen by running the MVGBench 3D-consistency metrics on SV3D variants (Tab. 3); CaP and ConvNextV2 are adopted because they give the best cPSNR/CD on GSO. The final comparison (Tab. 2) then uses the same MVGBench metrics on the same GSO/CO3D sets to report 'Ours' as best. No held-out benchmark or independent 3D-consistency test is used after the architecture is selected. The 'outperforms all' claim is therefore the selection objective reported as a result, rather than an out-of-sample prediction; the ranking is partly forced by the construction of the comparison.

full rationale

The core benchmark metrics are not circular by construction: the 3D self-consistency metric is defined as a discrepancy between two 3DGS reconstructions from disjoint view sets, and its behavior is independently validated on perfectly consistent ground-truth views (Tab. 1) and the VLM/oFID metrics are validated against user studies (Fig. 5, Sec. 4.1). The self-citations ([70], [71]) only motivate that 3D consistency matters and are not load-bearing. The main circularity is partial and confined to the method claim: MVGBench is used to select ViFiGen's components (Sec. 4.3) and the same MVGBench is then used to declare ViFiGen superior (Tab. 2), so the method result lacks an independent out-of-benchmark check. The manuscript's own Limitations section states that 'the 3DGS fitting is very inaccurate and fitting error instead of multi-view inconsistency dominates our consistency scores' for few-view methods, which is a validity risk for the metric on inconsistent generated views but not an equation-level circularity. Overall, the benchmark's central content is externally grounded, so the analysis is not fully circular; the score reflects the self-referential design-to-validation loop.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The benchmark's central measurement rests on several domain assumptions that are partially validated but not exhaustively tested. The free parameters are evaluation-protocol choices that affect comparability but are not fitted to game the result. No new physical or conceptual entities are introduced; ViFiGen is a method, not an entity.

free parameters (4)
  • Test-view sampling ranges for best-setup and robustness evaluation = elevation [-15, 45], azimuth [0, 360], camera distance [1.5, 1.9], FOV 42 degrees
    Chosen by hand in Supp. Sec. 6.1 to avoid favoring any method; these ranges affect comparability of scores across methods and datasets.
  • Normalization maximum max_e_mvg in Eq. 5 = max over evaluated methods
    Used for robustness scores; normalized values are relative to the specific set of methods evaluated, so numbers are not absolute and may change if the method set changes.
  • View overlap indices for 16/18/21 output views = specific index lists in Supp. Sec. 6.1
    Chosen by hand to balance 3DGS fitting accuracy with disjointness of view subsets; affects the computed consistency scores.
  • ICP reference 3DGS for real-world alignment = one MVG chosen as reference (unspecified)
    For real images, 3DGS reconstructions from different methods are aligned via ICP with uniform scaling to a reference 3DGS; no sensitivity analysis is provided for the choice of reference method.
assumptions (4)
  • domain assumption 3DGS optimization from 10 to 20 views yields geometry accurate enough that the discrepancy between two fits is dominated by multi-view inconsistency rather than by fitting artifacts.
    This is the core assumption of the 3D consistency metric, stated in Sec. 3.1 and validated only on perfectly consistent GT views (Tab. 1), not on inconsistent generated views.
  • domain assumption The pretrained VLM (InternVL2.5 73B) is an accurate judge of image quality and semantic attributes for the tested objects.
    Used in Sec. 3.2 for IQ-vlm and semantic metrics; a user study on 400 images with 80 users supports correlation, but the VLM may have biases not captured by the study.
  • domain assumption Uniform-scaling ICP aligns 3DGS reconstructions from different methods on real images.
    Invoked in Sec. 3.1 for real-data evaluation; no sensitivity analysis is given for the reference method or alignment quality.
  • domain assumption Manual annotations of front view and elevation angle for CO3D and MVImgNet images are correct.
    Used in Sec. 3.3 for generalization evaluation; annotation errors would bias the input conditions for all methods.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MVGBench: Comprehensive Benchmark for Multi-view Generation Models." pith.science (2026). https://pith.science/paper/YFBNTXE5

@misc{pith2026250700006,
  author       = {Pith},
  title        = {Pith review of: MVGBench: Comprehensive Benchmark for Multi-view Generation Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YFBNTXE5}},
  note         = {Machine review of arXiv:2507.00006}
}
read the original abstract

We propose MVGBench, a comprehensive benchmark for multi-view image generation models (MVGs) that evaluates 3D consistency in geometry and texture, image quality, and semantics (using vision language models). Recently, MVGs have been the main driving force in 3D object creation. However, existing metrics compare generated images against ground truth target views, which is not suitable for generative tasks where multiple solutions exist while differing from ground truth. Furthermore, different MVGs are trained on different view angles, synthetic data and specific lightings -- robustness to these factors and generalization to real data are rarely evaluated thoroughly. Without a rigorous evaluation protocol, it is also unclear what design choices contribute to the progress of MVGs. MVGBench evaluates three different aspects: best setup performance, generalization to real data and robustness. Instead of comparing against ground truth, we introduce a novel 3D self-consistency metric which compares 3D reconstructions from disjoint generated multi-views. We systematically compare 12 existing MVGs on 4 different curated real and synthetic datasets. With our analysis, we identify important limitations of existing methods specially in terms of robustness and generalization, and we find the most critical design choices. Using the discovered best practices, we propose ViFiGen, a method that outperforms all evaluated MVGs on 3D consistency. Our code, model, and benchmark suite will be publicly released.

Figures

Figures reproduced from arXiv: 2507.00006 by the authors.

Figure 1
Figure 1. We present MVGBench, a comprehensive evaluation suite for multi-view image generation models (MVGs). We propose ten [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Comparison of classic pair-wise metrics and our met￾rics. Classic metrics compare generated images independently to paired ground truth views, which represent only one of many cor￾rect solutions in the ambiguous single-view generation task. In the example shown, despite the inconsistent generated multi-views, they assign a higher score to Zero123 [2] while our metrics cor￾rectly identify SV3D [54] as the more 3D con… view at source ↗
Figure 3
Figure 3. 3D consistency metrics for multi-view generation models. After prompting the model to generate multi-views at target camera poses, we split the output views and fit 3D Gaussian Splatting (3DGS) separately into two view sets. We measure the geometric and texture consistency between two 3DGSs as the 3D consistency of the multi-view generation model. masked out. Our evaluation suite consists of comprehensive evaluation… view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Example results on CO3D [45] (row 1), MVImgnet [75] (row 2), Omni3D [63] (row 3) and GSO [12] (row 4). Vivid123 [30] generates images that look good but are 3D inconsistent while SyncDreamer [38] images are 3D consistent but the image quality is worse. SV3D [54] and ou…
Figure 5
Figure 5. Figure 5: Validating vision language model (VLM) based met￾rics. Our VLM metrics strongly correlate with human perception (Pearson coefficient confidence interval: 0.95). of computing FID against the full dataset of all objects, we compute FID per object instance, and then avera…
Figure 6
Figure 6. Figure 6: Trade-off between 3D consistency and image quality. No method can achieve the best performance in both dimensions. quality, but are not 3D consistent. Conversely, methods like SyncDreamer [38], EscherNet [28] and MVDFusion [17] can generate 3D consistent images at the …
Figure 7
Figure 7. Figure 7: Robustness w.r.t different light intensity, azimuth and elevation angles. Some methods (EscherNet [28]) are sensitive to dark lighting while others (SyncDreamer [38]) are sensitive to strong lighting. Some methods (EscherNet, Vivid123 [30]) are also sensitive to the in…
Figure 8
Figure 8. Figure 8: Example question from our user study on the alignment [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 10
Figure 10. Figure 10: The most challenging test images from GSO[12], Omni3D[63], MVImgnet[75] and CO3D[45]. Methods produce the most 3D inconsistent images for these inputs due to their com￾plex geometric structure or high frequency details. that the common challenging images are the objec…
Figure 9
Figure 9. Figure 9: Example question from our user study on the alignment [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 11
Figure 11. Figure 11: Degradation of image quality after passing through the [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

83 extracted references · 56 canonical work pages

  1. [1]

    http://virtualhumans.mpi-inf.mpg.de/people.html. 9

  2. [2]

    Stable zero123: Quality 3d object generation from single images.https://stability.ai/news/ stable-zero123-3d-generation, 2023

    Stability AI. Stable zero123: Quality 3d object generation from single images.https://stability.ai/news/ stable-zero123-3d-generation, 2023. 2, 3, 4, 5

    2023

  3. [3]

    Met3r: Measur- ing multi-view consistency in generated images

    Mohammad Asim, Christopher Wewer, Thomas Wimmer, Bernt Schiele, and Jan Eric Lenssen. Met3r: Measur- ing multi-view consistency in generated images. InarXiv preprint, 2025. 3

    2025

  4. [4]

    Ten years of pedestrian detection, what have we learned?, 2014

    Rodrigo Benenson, Mohamed Omran, Jan Hosang, and Bernt Schiele. Ten years of pedestrian detection, what have we learned?, 2014. 3

    work page 2014

  5. [5]

    Behave: Dataset and method for tracking human object inter- actions

    Bharat Lal Bhatnagar, Xianghui Xie, Ilya Petrov, Cristian Sminchisescu, Christian Theobalt, and Gerard Pons-Moll. Behave: Dataset and method for tracking human object inter- actions. InIEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022. 3

    work page 2022

  6. [6]

    Align your latents: High-resolution video synthesis with la- tent diffusion models

    Andreas Blattmann, Robin Rombach, Huan Ling, Tim Dock- horn, Seung Wook Kim, Sanja Fidler, and Karsten Kreis. Align your latents: High-resolution video synthesis with la- tent diffusion models. InIEEE Conference on Computer Vi- sion and Pattern Recognition (CVPR), 2023. 3

    work page 2023

  7. [7]

    Muse: Text-to-image generation via masked generative transformers.arXiv preprint arXiv:2301.00704,

    Jonathan Chang, Barret Zoph, Andrew Dai, Zihang Chen, and Quoc V Le. Muse: Text-to-image generation via masked generative transformers.arXiv preprint arXiv:2301.00704,

    arXiv

  8. [8]

    Expanding performance boundaries of open-source multimodal models with model, data, and test- time scaling.arXiv preprint arXiv:2412.05271, 2024

    Zhe Chen, Weiyun Wang, Yue Cao, Yangzhou Liu, Zhang- wei Gao, Erfei Cui, Jinguo Zhu, Shenglong Ye, Hao Tian, Zhaoyang Liu, et al. Expanding performance boundaries of open-source multimodal models with model, data, and test- time scaling.arXiv preprint arXiv:2412.05271, 2024. 5, 6, 2

    arXiv 2024

Show all 83 references
  1. [9]

    V3d: Video diffusion models are effective 3d generators.arXiv preprint arXiv:2403.06738, 2024

    Zilong Chen, Yikai Wang, Feng Wang, Zhengyi Wang, and Huaping Liu. V3d: Video diffusion models are effective 3d generators.arXiv preprint arXiv:2403.06738, 2024. 3, 6, 2, 4, 5

    2024 arXiv

  2. [10]

    Llamagen: Autoregressive large language models for visual image generation.arXiv preprint arXiv:2406.06525, 2024

    Zihang Chen, Ming Zhao, et al. Llamagen: Autoregressive large language models for visual image generation.arXiv preprint arXiv:2406.06525, 2024. 2

    2024 arXiv

  3. [11]

    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 V oleti, Samir Yitzhak Gadre, Eli VanderBilt, Anirud- dha Kembhavi, Carl V ondrick, Georgia Gkioxari, Kiana Ehsani, Ludwig Schmidt, and Ali Farhadi. Obja...

    2023

  4. [12]

    McHugh, and Vincent Vanhoucke

    Laura Downs, Anthony Francis, Nate Koenig, Brandon Kin- man, Ryan Hickman, Krista Reymann, Thomas B. McHugh, and Vincent Vanhoucke. Google scanned objects: A high- quality dataset of 3d scanned household items. InProceed- ings of the IEEE International Conference on Robotics a...

    2022

  5. [13]

    Stable-edit: Text-based real image editing with stable diffusion models.https://github.com/ feizc/Stable-Edit, 2023

    Zongcai Fei. Stable-edit: Text-based real image editing with stable diffusion models.https://github.com/ feizc/Stable-Edit, 2023. 2 9

    2023

  6. [14]

    Barron, and Ben Poole

    Ruiqi Gao, Aleksander Hoły ´nski, 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. 3

    2024 arXiv

  7. [15]

    Barron, and Ben Poole

    Ruiqi Gao, Aleksander Hoły ´nski, Philipp Henzler, Arthur Brussee, Ricardo Martin-Brualla, Pratul Srinivasan, Jonathan T. Barron, and Ben Poole. Mvreward: Better aligning and evaluating multi-view diffusion models with human preferences.arXiv preprint arXiv:2412.06614, 2024. 3

    2024 arXiv

  8. [16]

    Bop: Benchmark for 6d object pose esti- mation, 2018

    Tomas Hodan, Frank Michel, Eric Brachmann, Wadim Kehl, Anders Glent Buch, Dirk Kraft, Bertram Drost, Joel Vidal, Stephan Ihrke, Xenophon Zabulis, Caner Sahin, Fabian Man- hardt, Federico Tombari, Tae-Kyun Kim, Jiri Matas, and Carsten Rother. Bop: Benchmark for 6d object pose e...

    2018

  9. [17]

    MVD-Fusion: Single-view 3D via Depth- consistent Multi-view Generation

    Hanzhe Hu, Zhizhuo Zhou, Varun Jampani, and Shub- ham Tulsiani. MVD-Fusion: Single-view 3D via Depth- consistent Multi-view Generation. 3, 5, 6, 7, 4

  10. [18]

    Turbo3d: Ultra-fast text-to-3d generation, 2024

    Hanzhe Hu, Tianwei Yin, Fujun Luan, Yiwei Hu, Hao Tan, Zexiang Xu, Sai Bi, Shubham Tulsiani, and Kai Zhang. Turbo3d: Ultra-fast text-to-3d generation, 2024. 3

    2024

  11. [19]

    2d gaussian splatting for geometrically accu- rate radiance fields

    Binbin Huang, Zehao Yu, Anpei Chen, Andreas Geiger, and Shenghua Gao. 2d gaussian splatting for geometrically accu- rate radiance fields. InSIGGRAPH 2024 Conference Papers. Association for Computing Machinery, 2024. 1

    2024

  12. [20]

    TeCH: Text-guided Reconstruction of Lifelike Clothed Humans

    Yangyi Huang, Hongwei Yi, Yuliang Xiu, Tingting Liao, Ji- axiang Tang, Deng Cai, and Justus Thies. TeCH: Text-guided Reconstruction of Lifelike Clothed Humans. InInternational Conference on 3D Vision (3DV), 2024. 2

    2024

  13. [21]

    EpiDiff: Enhancing Multi- View Synthesis via Localized Epipolar-Constrained Diffu- sion

    Zehuan Huang, Hao Wen, Junting Dong, Yaohui Wang, Yangguang Li, Xinyuan Chen, Yan-Pei Cao, Ding Liang, Yu Qiao, Bo Dai, and Lu Sheng. EpiDiff: Enhancing Multi- View Synthesis via Localized Epipolar-Constrained Diffu- sion. 3, 6, 4, 5

  14. [22]

    Mv-adapter: Multi-view consistent image generation made easy.arXiv preprint arXiv:2412.03632, 2024

    Zehuan Huang, Yuanchen Guo, Haoran Wang, Ran Yi, Lizhuang Ma, Yan-Pei Cao, and Lu Sheng. Mv-adapter: Multi-view consistent image generation made easy.arXiv preprint arXiv:2412.03632, 2024. 3

    2024 arXiv

  15. [23]

    Vbench: Comprehensive benchmark suite for video generative models

    Ziqi Huang, Yilun Sheng, Yujian Liu, Yujie Lu, and Wenhu Chen. Vbench: Comprehensive benchmark suite for video generative models. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR),

  16. [24]

    Re- thinking fid: Towards a better evaluation metric for image generation

    Sadeep Jayasumana, Srikumar Ramalingam, Andreas Veit, Daniel Glasner, Ayan Chakrabarti, and Sanjiv Kumar. Re- thinking fid: Towards a better evaluation metric for image generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), page...

    2024

  17. [25]

    NVS-Adapter: Plug-and-Play Novel View Syn- thesis from a Single Image, 2024

    Yoonwoo Jeong, Jinwoo Lee, Chiheon Kim, Minsu Cho, and Doyup Lee. NVS-Adapter: Plug-and-Play Novel View Syn- thesis from a Single Image, 2024. 3

    2024

  18. [26]

    SPAD: Spatially Aware Multi-View Diffusers

    Yash Kant, Aliaksandr Siarohin, Ziyi Wu, Michael Vasilkovsky, Guocheng Qian, Jian Ren, Riza Alp Guler, Bernard Ghanem, Sergey Tulyakov, and Igor Gilitschenski. SPAD: Spatially Aware Multi-View Diffusers. 2, 3

  19. [27]

    3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics, 42 (4), 2023

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics, 42 (4), 2023. 3, 1

    2023

  20. [28]

    EscherNet: A Generative Model for Scalable View Synthesis

    Xin Kong, Shikun Liu, Xiaoyang Lyu, Marwan Taher, Xiao- juan Qi, and Andrew J Davison. EscherNet: A Generative Model for Scalable View Synthesis. 2, 3, 5, 6, 7, 8, 4

  21. [29]

    Imagenhub: Standardiz- ing the evaluation of conditional image generation models

    Max Ku, Tianle Li, Kai Zhang, Yujie Lu, Xingyu Fu, Wen- wen Zhuang, and Wenhu Chen. Imagenhub: Standardiz- ing the evaluation of conditional image generation models. InProceedings of the Twelfth International Conference on Learning Representations (ICLR), 2024. 3

    2024

  22. [30]

    ViVid-1-to-3: Novel View Synthesis with Video Diffusion Models

    Jeong-gi Kwak, Erqun Dong, Yuhe Jin, Hanseok Ko, Shweta Mahajan, and Kwang Moo Yi. ViVid-1-to-3: Novel View Synthesis with Video Diffusion Models. 3, 5, 6, 8, 4

  23. [31]

    Nvcomposer: Boosting generative novel view synthesis with multiple sparse and unposed images, 2024

    Lingen Li, Zhaoyang Zhang, Yaowei Li, Jiale Xu, Wenbo Hu, Xiaoyu Li, Weihao Cheng, Jinwei Gu, Tianfan Xue, and Ying Shan. Nvcomposer: Boosting generative novel view synthesis with multiple sparse and unposed images, 2024. 3

    2024

  24. [32]

    Era3d: High-resolution multiview diffusion using efficient row-wise attention.arXiv preprint arXiv:2405.11616, 2024

    Peng Li, Yuan Liu, Xiaoxiao Long, Feihu Zhang, Cheng Lin, Mengfei Li, Xingqun Qi, Shanghang Zhang, Wenhan Luo, Ping Tan, et al. Era3d: High-resolution multiview diffusion using efficient row-wise attention.arXiv preprint arXiv:2405.11616, 2024. 3

    2024 arXiv

  25. [33]

    Evaluation of text-to-video generation models: A dy- namics perspective.Advances in Neural Information Pro- cessing Systems, 37:109790–109816, 2024

    Mingxiang Liao, Qixiang Ye, Wangmeng Zuo, Fang Wan, Tianyu Wang, Yuzhong Zhao, Jingdong Wang, Xinyu Zhang, et al. Evaluation of text-to-video generation models: A dy- namics perspective.Advances in Neural Information Pro- cessing Systems, 37:109790–109816, 2024. 3

    2024

  26. [34]

    Tingting Liao, Hongwei Yi, Yuliang Xiu, Jiaxiang Tang, Yangyi Huang, Justus Thies, and Michael J. Black. TADA! Text to Animatable Digital Avatars. InInternational Confer- ence on 3D Vision (3DV), 2024. 2

    2024

  27. [35]

    Stylerf: Zero-shot 3d style transfer of neural radiance fields

    Kunhao Liu, Fangneng Zhan, Yiwen Chen, Jiahui Zhang, Yingchen Yu, Abdulmotaleb El Saddik, Shijian Lu, and Eric Xing. Stylerf: Zero-shot 3d style transfer of neural radiance fields. 2023. 3

    2023

  28. [36]

    One-2-3-45: Any Single Im- age to 3D Mesh in 45 Seconds without Per-Shape Optimiza- tion

    Minghua Liu, Chao Xu, Haian Jin, Linghao Chen, Mukund T, Zexiang Xu, and Hao Su. One-2-3-45: Any Single Im- age to 3D Mesh in 45 Seconds without Per-Shape Optimiza- tion. InAnnual Conference on Neural Information Process- ing Systems (NeurIPS), 2023. 2, 3

    2023

  29. [37]

    Zero-1-to-3: Zero-shot one image to 3d object, 2023

    Ruoshi Liu, Rundi Wu, Basile Van Hoorick, Pavel Tok- makov, Sergey Zakharov, and Carl V ondrick. Zero-1-to-3: Zero-shot one image to 3d object, 2023. 2, 3, 6, 7

    2023

  30. [38]

    Syncdreamer: Gen- erating multiview-consistent images from a single-view im- age.arXiv preprint arXiv:2309.03453, 2023

    Yuan Liu, Cheng Lin, Zijiao Zeng, Xiaoxiao Long, Lingjie Liu, Taku Komura, and Wenping Wang. Syncdreamer: Gen- erating multiview-consistent images from a single-view im- age.arXiv preprint arXiv:2309.03453, 2023. 2, 3, 5, 6, 7, 8, 4

    2023 arXiv

  31. [39]

    Wonder3d: Sin- gle image to 3d using cross-domain diffusion.arXiv preprint arXiv:2310.15008, 2023

    Xiaoxiao Long, Yuan-Chen Guo, Cheng Lin, Yuan Liu, Zhiyang Dou, Lingjie Liu, Yuexin Ma, Song-Hai Zhang, Marc Habermann, Christian Theobalt, et al. Wonder3d: Sin- gle image to 3d using cross-domain diffusion.arXiv preprint arXiv:2310.15008, 2023. 3, 6

    arXiv

  32. [40]

    Direct2.5: Diverse text-to-3d generation via multi-view 2.5d diffusion.Computer Vision and Pattern Recognition (CVPR), 2024

    Yuanxun Lu, Jingyang Zhang, Shiwei Li, Tian Fang, David McKinnon, Yanghai Tsin, Long Quan, Xun Cao, and Yao 10 Yao. Direct2.5: Diverse text-to-3d generation via multi-view 2.5d diffusion.Computer Vision and Pattern Recognition (CVPR), 2024. 2, 3

    2024

  33. [41]

    Taming 3dgs: High-quality radiance fields with limited resources

    Saswat Subhajyoti Mallick, Rahul Goel, Bernhard Kerbl, Markus Steinberger, Francisco Vicente Carrasco, and Fer- nando De La Torre. Taming 3dgs: High-quality radiance fields with limited resources. InSIGGRAPH Asia 2024 Con- ference Papers, New York, NY , USA, 2024. Association ...

    2024

  34. [42]

    Dinov2: Learning robust visual features with- out supervision, 2024

    Maxime Oquab, Timoth ´ee Darcet, Th ´eo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, Mah- moud Assran, Nicolas Ballas, Wojciech Galuba, Russell Howes, Po-Yao Huang, Shang-Wen Li, Ishan Misra, Michae...

    2024

  35. [43]

    Barron, and Ben Milden- hall

    Ben Poole, Ajay Jain, Jonathan T. Barron, and Ben Milden- hall. Dreamfusion: Text-to-3d using 2d diffusion.arXiv,

  36. [44]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision. InProceedings of the...

    2021

  37. [45]

    Com- mon objects in 3d: Large-scale learning and evaluation of real-life 3d category reconstruction

    Jeremy Reizenstein, Roman Shapovalov, Philipp Henzler, Luca Sbordone, Patrick Labatut, and David Novotny. Com- mon objects in 3d: Large-scale learning and evaluation of real-life 3d category reconstruction. InProceedings of the IEEE/CVF Conference on Computer Vision and Patter...

    2021

  38. [46]

    High-resolution image syn- thesis with latent diffusion models, 2021

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj¨orn Ommer. High-resolution image syn- thesis with latent diffusion models, 2021. 2, 3, 7

    2021

  39. [47]

    Stable diffusion 3.5: High- resolution image synthesis with latent diffusion models.Sta- bility AI, 2024

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj¨orn Ommer. Stable diffusion 3.5: High- resolution image synthesis with latent diffusion models.Sta- bility AI, 2024. 2, 3

    2024

  40. [48]

    Zero123++: a single image to consistent multi-view dif- fusion base model, 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 dif- fusion base model, 2023. 2, 3, 6

    2023

  41. [49]

    Mvdream: Multi-view diffusion for 3d gen- eration.arXiv:2308.16512, 2023

    Yichun Shi, Peng Wang, Jianglong Ye, Long Mai, Kejie Li, and Xiao Yang. Mvdream: Multi-view diffusion for 3d gen- eration.arXiv:2308.16512, 2023. 2, 3

    2023 arXiv

  42. [50]

    Dragdiffusion: Harnessing diffu- sion models for interactive point-based image editing.arXiv preprint arXiv:2306.14435, 2023

    Yujun Shi, Chuhui Xue, Jiachun Pan, Wenqing Zhang, Vin- cent YF Tan, and Song Bai. Dragdiffusion: Harnessing diffu- sion models for interactive point-based image editing.arXiv preprint arXiv:2306.14435, 2023. 2

    2023 arXiv

  43. [51]

    Deqing Sun, Stefan Roth, and Michael J. Black. Secrets of optical flow estimation and their principles. In2010 IEEE Computer Society Conference on Computer Vision and Pat- tern Recognition, pages 2432–2439, 2010. 3

    2010

  44. [52]

    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. 3, 4, 2

    2024 arXiv

  45. [53]

    Visual autoregressive modeling: Scalable image generation via next-scale prediction

    Keyu Tian, Yi Jiang, Zehuan Yuan, Bingyue Peng, and Li- wei Wang. Visual autoregressive modeling: Scalable image generation via next-scale prediction. 2024. 2

    2024

  46. [54]

    SV3D: Novel Multi-view Synthesis and 3D Generation from a Single Image using La- tent Video Diffusion, 2024

    Vikram V oleti, Chun-Han Yao, Mark Boss, Adam Letts, David Pankratz, Dmitry Tochilkin, Christian Laforte, Robin Rombach, and Varun Jampani. SV3D: Novel Multi-view Synthesis and 3D Generation from a Single Image using La- tent Video Diffusion, 2024. 2, 3, 5, 6, 7, 8, 4

    2024

  47. [55]

    Exploiting diffusion prior for real-world image super-resolution.arXiv preprint arXiv:2305.07015, 2023

    Jianyi Wang, Zongsheng Yue, Shangchen Zhou, Kelvin CK Chan, and Chen Change Loy. Exploiting diffusion prior for real-world image super-resolution.arXiv preprint arXiv:2305.07015, 2023. 2

    2023 arXiv

  48. [56]

    Imagedream: Image-prompt multi-view diffusion for 3d generation.arXiv preprint arXiv:2312.02201, 2023

    Peng Wang and Yichun Shi. Imagedream: Image-prompt multi-view diffusion for 3d generation.arXiv preprint arXiv:2312.02201, 2023. 2, 3, 6

    2023 arXiv

  49. [57]

    Prolificdreamer: High-fidelity and diverse text-to-3d generation with variational score distilla- tion.arXiv preprint arXiv:2305.16213, 2023

    Zhengyi Wang, Cheng Lu, Yikai Wang, Fan Bao, Chongxuan Li, Hang Su, and Jun Zhu. Prolificdreamer: High-fidelity and diverse text-to-3d generation with variational score distilla- tion.arXiv preprint arXiv:2305.16213, 2023. 2

    2023 arXiv

  50. [58]

    Novel View Synthesis with Diffusion Models,

    Daniel Watson, William Chan, Ricardo Martin-Brualla, Jonathan Ho, Andrea Tagliasacchi, and Mohammad Norouzi. Novel View Synthesis with Diffusion Models,

  51. [59]

    Con- vnext v2: Co-designing and scaling convnets with masked autoencoders.arXiv preprint arXiv:2301.00808, 2023

    Sanghyun Woo, Shoubhik Debnath, Ronghang Hu, Xinlei Chen, Zhuang Liu, In So Kweon, and Saining Xie. Con- vnext v2: Co-designing and scaling convnets with masked autoencoders.arXiv preprint arXiv:2301.00808, 2023. 3, 5, 7

    2023 arXiv

  52. [60]

    Direct and explicit 3d generation from a single image, 2024

    Haoyu Wu, Meher Gitika Karumuri, Chuhang Zou, Seung- bae Bang, Yuelong Li, Dimitris Samaras, and Sunil Hadap. Direct and explicit 3d generation from a single image, 2024. 2

    2024

  53. [61]

    Direct and Explicit 3D Generation from a Single Image,

    Haoyu Wu, Meher Gitika Karumuri, Chuhang Zou, Seung- bae Bang, Yuelong Li, Dimitris Samaras, and Sunil Hadap. Direct and Explicit 3D Generation from a Single Image,

  54. [62]

    One-step effective diffusion network for real-world image super-resolution.arXiv preprint arXiv:2406.08177, 2024

    Rongyuan Wu, Lingchen Sun, Zhiyuan Ma, and Lei Zhang. One-step effective diffusion network for real-world image super-resolution.arXiv preprint arXiv:2406.08177, 2024. 2

    2024 arXiv

  55. [63]

    Omniobject3d: Large-vocabulary 3d object dataset for realistic perception, reconstruction and generation

    Tong Wu, Jiarui Zhang, Xiao Fu, Yuxin Wang, Liang Pan Jiawei Ren, Wayne Wu, Lei Yang, Jiaqi Wang, Chen Qian, Dahua Lin, and Ziwei Liu. Omniobject3d: Large-vocabulary 3d object dataset for realistic perception, reconstruction and generation. InIEEE/CVF Conference on Computer Vi...

    2023

  56. [64]

    Gpt- 4v(ision) is a human-aligned evaluator for text-to-3d genera- tion

    Tong Wu, Guandao Yang, Zhibing Li, Kai Zhang, Ziwei Liu, Leonidas Guibas, Dahua Lin, and Gordon Wetzstein. Gpt- 4v(ision) is a human-aligned evaluator for text-to-3d genera- tion. InCVPR, 2024. 3

    2024

  57. [65]

    Xianghui Xie, Xi Wang, Nikos Athanasiou, Bharat Lal Bhat- nagar, Chun-Hao P. Huang, Kaichun Mo, Hao Chen, Xia Jia, Zerui Zhang, Liangxian Cui, Xiao Lin, Bingqiao Qian, Jie Xiao, Wenfei Yang, Hyeongjin Nam, Daniel Sungho Jung, Kihoon Kim, Kyoung Mu Lee, Otmar Hilliges, and Gera...

    2024

  58. [66]

    Dynamicrafter: Animating open-domain images with video diffusion priors.arXiv preprint arXiv:2310.12190, 2023

    Jinbo Xing, Menghan Xia, Yong Zhang, Haoxin Chen, Wangbo Yu, Hanyuan Liu, Xintao Wang, Tien-Tsin Wong, and Ying Shan. Dynamicrafter: Animating open-domain images with video diffusion priors.arXiv preprint arXiv:2310.12190, 2023. 3

    arXiv

  59. [67]

    CamCo: Camera- Controllable 3D-Consistent Image-to-Video Generation,

    Dejia Xu, Weili Nie, Chao Liu, Sifei Liu, Jan Kautz, Zhangyang Wang, and Arash Vahdat. CamCo: Camera- Controllable 3D-Consistent Image-to-Video Generation,

  60. [68]

    Imagere- ward: Learning and evaluating human preferences for text- to-image generation

    Jiazheng Xu, Yuxuan Du, Yilun Zhao, Yifan Xu, Lei Li, Maosong Sun, Zhiyuan Liu, and Yang Liu. Imagere- ward: Learning and evaluating human preferences for text- to-image generation. InAdvances in Neural Information Processing Systems (NeurIPS), 2023. 3

    2023

  61. [69]

    Instantmesh: Efficient 3d mesh generation from a single image with sparse-view large reconstruction models.arXiv preprint arXiv:2404.07191,

    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 preprint arXiv:2404.07191,

    arXiv

  62. [70]

    Pons-Moll

    Yuxuan Xue, Xianghui Xie, Riccardo Marin, and Gerard. Pons-Moll. Gen-3Diffusion: Realistic Image-to-3D Genera- tion via 2D & 3D Diffusion Synergy. 2024. 2, 3

    2024

  63. [71]

    Human 3diffusion: Realistic avatar creation via explicit 3d consistent diffusion models

    Yuxuan Xue, Xianghui Xie, Riccardo Marin, and Gerard Pons-Moll. Human 3diffusion: Realistic avatar creation via explicit 3d consistent diffusion models. InArxiv, 2024. 2, 3

    2024

  64. [72]

    Hi3d: Pursuing high- resolution image-to-3d generation with video diffusion mod- els

    Haibo Yang, Yang Chen, Yingwei Pan, Ting Yao, Zhineng Chen, Chong-Wah Ngo, and Tao Mei. Hi3d: Pursuing high- resolution image-to-3d generation with video diffusion mod- els. InACM MM, 2024. 3, 6, 4, 5

    2024

  65. [73]

    Pixel-aware stable diffusion for realistic image super-resolution and personalized stylization.arXiv preprint arXiv:2308.14469, 2023

    Tao Yang, Rongyuan Wu, Peiran Ren, Xuansong Xie, and Lei Zhang. Pixel-aware stable diffusion for realistic image super-resolution and personalized stylization.arXiv preprint arXiv:2308.14469, 2023. 2

    2023 arXiv

  66. [74]

    ViewFusion: Towards Multi-View Consistency via Interpolated Denoising

    Xianghui Yang, Yan Zuo, Sameera Ramasinghe, Loris Baz- zani, and Gil Avraham. ViewFusion: Towards Multi-View Consistency via Interpolated Denoising. 6, 3, 4, 5

  67. [75]

    Mvimgnet: A large-scale dataset of multi-view images

    Xianggang Yu, Mutian Xu, Yidan Zhang, Haolin Liu, Chongjie Ye, Yushuang Wu, Zizheng Yan, Tianyou Liang, Guanying Chen, Shuguang Cui, and Xiaoguang Han. Mvimgnet: A large-scale dataset of multi-view images. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Patt...

    2023

  68. [76]

    Mip-splatting: Alias-free 3d gaussian splat- ting

    Zehao Yu, Anpei Chen, Binbin Huang, Torsten Sattler, and Andreas Geiger. Mip-splatting: Alias-free 3d gaussian splat- ting. InProceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pages 19447– 19456, 2024. 1

    2024

  69. [77]

    Gaussian opacity fields: Efficient high-quality compact surface recon- struction in unbounded scenes.arXiv:2404.10772, 2024

    Zehao Yu, Torsten Sattler, and Andreas Geiger. Gaussian opacity fields: Efficient high-quality compact surface recon- struction in unbounded scenes.arXiv:2404.10772, 2024. 1

    2024 arXiv

  70. [78]

    Adding conditional control to text-to-image diffusion models.arXiv preprint arXiv:2302.05543, 2023

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models.arXiv preprint arXiv:2302.05543, 2023. 2

    2023 arXiv

  71. [79]

    Diffcollage: Parallel generation of large content with diffusion models

    Qinsheng Zhang, Jiaming Song, Xun Huang, Yongxin Chen, and Ming yu Liu. Diffcollage: Parallel generation of large content with diffusion models. InCVPR, 2023. 2

    2023

  72. [80]

    Free3D: Consis- tent Novel View Synthesis Without 3D Representation

    Chuanxia Zheng and Andrea Vedaldi. Free3D: Consis- tent Novel View Synthesis Without 3D Representation. In 2024 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), pages 9720–9731, Seattle, W A, USA, 2024. IEEE. 3, 6, 7, 4, 5 12 MVGBench: a Comprehensive B...

    2024

  73. [81]

    What is the main color(s) of this object? simply answer the color(s), summarize to less than 4 colors

    Implementation Details We discuss the details of our metrics and experiment se- tups. Our benchmark suite and pre-trained models will be publicly released. 6.1. MVGBench Metric Implementation View sets split.For 3D consistency metric, we split the generated multi-view images i...

  74. [82]

    6 (GSO [12]), Tab

    Additional Results and Analysis Full evaluation results.We show all scores of our MVG- Bench from all evaluated methods on four datasets in Tab. 6 (GSO [12]), Tab. 7 (Omni3D [63]), Tab. 8 (CO3D [45]), and Tab. 9 (MVImgnet [75]). It can be seen that our method achieves the best...

  75. [83]

    Despite robust to various settings, there are still limita- tions of our benchmark

    Limitations and Future Work We present the first comprehensive benchmark to evalu- ate 3D consistency of object multi-view generation mod- els. Despite robust to various settings, there are still limita- tions of our benchmark. First, our method cannot evaluate methods that ge...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.