Pith. sign in

REVIEW 2 major objections 5 minor 69 references

StudioRecon rebuilds high-fidelity 4D human scenes from only a handful of low-overlap cameras by giving backgrounds and people different priors.

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

2026-07-13 05:12 UTC pith:ZS6YVNWR

load-bearing objection Solid engineering pipeline that actually delivers SOTA low-overlap multi-human 4D recon by cleanly decoupling diffusion backgrounds from SMPL humans; the GEN3C dependence is real but not a load-bearing hole. the 2 major comments →

arxiv 2607.09125 v1 pith:ZS6YVNWR submitted 2026-07-10 cs.CV

4D Human-Scene Reconstruction from Low-Overlap Captures

Minhyuk Hwang , Sangmin Kim , Seunguk Do , Daneul Kim , Jaesik Park This is my paper
classification cs.CV
keywords 4D reconstructionGaussian splattingsparse-view capturehuman-scene reconstructionvideo diffusionSMPLnovel view synthesislow-overlap cameras
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.

Dense camera arrays already deliver good volumetric human capture, but real gyms, homes, and sports venues usually offer only a few cameras that barely share views, so large regions stay unobserved and joint 4D methods leave visible artifacts. StudioRecon argues that backgrounds and humans need different priors: a camera-controlled video diffusion model can invent hundreds of dense novel views of the static scene for photometric supervision, while parametric body models supply geometric constraints that let deformable Gaussian humans be initialized and tracked even under occlusion and sparse overlap. After the two parts are reconstructed separately, a recursive single-step diffusion enhancer injects motion-adaptive optical-flow consistency so the composite stays sharp and free of flicker. Across four real multi-person datasets the pipeline reports the best novel-view numbers and also supports free camera paths and actor replacement from the same reconstruction.

Core claim

When cameras are few and largely non-overlapping, jointly representing humans and background in one dynamic model entangles their errors; decoupling them lets video diffusion densify background supervision while multi-view keypoint triangulation and SMPL fitting robustly initialize deformable human Gaussians, after which motion-adaptive recursive diffusion enhancement removes residual artifacts and restores temporal coherence in free-viewpoint renderings.

What carries the argument

Decoupled Gaussian reconstruction with complementary priors: video diffusion synthesizes dense camera-controlled novel views to supervise static background Gaussians, while cross-view spatial-and-pose affinity association plus triangulated keypoint fitting initialize SMPL-skinned human Gaussians; a recursive enhancement module then blends optical-flow-warped prior outputs into single-step diffusion so the composite is cleaned without flicker.

Load-bearing premise

The method assumes that a camera-controlled video diffusion model can invent hundreds of geometrically trustworthy novel views of the static background that, once humans are masked, supply reliable photometric supervision without uncorrectable inconsistencies.

What would settle it

Replace the synthesized background views with pure noise or with deliberately multi-view-inconsistent images, retrain on the same four-camera splits of the paper's datasets, and check whether held-out novel-view PSNR and LPIPS fall back to or below the joint baselines; collapse would falsify the densification claim.

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

If this is right

  • Four roughly 90°-spaced cameras become enough for usable free-viewpoint 4D capture of multi-person scenes in sports, healthcare, and home settings.
  • The same reconstructed Gaussians support novel trajectories such as dolly zoom or lateral oscillation without extra capture.
  • Actor replacement can be performed by swapping the human Gaussians and a single edited reference while leaving the background untouched.
  • Joint sparse-view methods that keep people and background in one representation will continue to leave under-observed artifacts that densified, decoupled supervision can avoid.

Where Pith is reading between the lines

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

  • The same split-prior idea may transfer to dynamic objects that lack parametric body models, such as balls or handheld props, if a suitable geometric prior can be supplied.
  • Once camera-controlled diffusion becomes cheaper, the offline bottleneck of view synthesis could shrink enough for near-interactive in-the-wild studio capture.
  • Baking shadows into the static background at the first frame is a structural gap; treating shadows as a second dynamic layer would be a direct test of how far decoupling can be pushed.

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 / 5 minor

Summary. StudioRecon reconstructs dynamic multi-human scenes from sparse, low-overlap multi-view video (typically four cameras ~90° apart) by decoupling static background from deformable humans. Background Gaussians are densified with hundreds of camera-controlled novel views synthesized by GEN3C video diffusion (Sec. 3.1), humans are initialized via geometry-driven cross-view identity association and multi-view keypoint triangulation into SMPL (Sec. 3.2, Eqs. 1–4), and the two are optimized separately before a recursive single-step diffusion enhancer with motion-adaptive optical-flow consistency injection (Sec. 3.4, Eq. 7). The method reports state-of-the-art novel-view PSNR/SSIM/LPIPS on four real-world datasets (EgoHumans, Harmony4D, Mobile Stage, SelfCap) against Dyn-3DGS, MonoFusion and STG, with ablations isolating view synthesis, association strategy, enhancement and noise sensitivity, plus applications in free-trajectory rendering and human replacement.

Significance. If the results hold, the work supplies a practical, modular pipeline that brings high-fidelity 4D human-scene capture into the low-overlap “in-the-wild studio” regime that dense volumetric systems cannot reach. The explicit decoupling of complementary priors (diffusion for static background, SMPL for articulated humans) is a clean engineering insight, the multi-view association/triangulation module is carefully validated (97.8 % accuracy), and the motion-adaptive consistency injection measurably reduces temporal flicker. Strong empirical support—four diverse datasets, three competitive baselines, component ablations, seen/unseen fidelity splits, and noise-sensitivity tests—makes the SOTA claim credible and the applications immediately useful for free-viewpoint video and virtual production.

major comments (2)
  1. The largest quantitative gain (+2.4 dB PSNR, Table 4) is attributed to GEN3C dense view synthesis (Sec. 3.1, L=481). While Appendix G correctly shows GEN3C is geometrically unreliable for humans, residual human contamination, depth misalignment or view-inconsistent background texture can still leak into the frozen background Gaussians (Eq. 5). The 21 px mask dilation, mid-optimization Difix3D+ refinement and recursive enhancer only partially address this; a direct quantitative audit of residual human pixels or multi-view photometric inconsistency on the synthesized background views (before masking) would strengthen the claim that the densified supervision is sufficiently clean.
  2. Table S5 shows that even 30 imes denser GEN3C supervision still leaves baselines well below the full pipeline on foreground metrics. This is useful, yet the paper never reports an ablation that freezes the background Gaussians obtained from GEN3C and replaces only the human module (or vice versa). Without that isolation it remains unclear how much of the final SOTA margin is truly due to the SMPL-guided human path versus residual background quality, which is load-bearing for the central “decoupled priors” claim.
minor comments (5)
  1. No error bars or multi-seed statistics accompany the PSNR/SSIM/LPIPS tables; given the stochastic nature of diffusion synthesis and enhancement, at least a short note on run-to-run variance would help.
  2. Affinity weights (w_p=0.9, w_θ=0.1) and the 0.3 m distance threshold (Sec. 3.2.2) are stated without sensitivity analysis beyond the binary “only spatial / only pose” ablation; a short sweep would confirm robustness.
  3. Limitations (Sec. 4.5, Appendix I) correctly note missing dynamic objects and baked shadows; a brief quantitative measure of how often these artifacts appear on the four datasets would make the scope clearer.
  4. Figure 2 overview is dense; a short textual walk-through of data flow between the four stages would improve readability for readers new to the pipeline.
  5. Runtime (Appendix F) is reported only for A6000; a note on memory footprint for the 481-view background optimization would aid reproducibility.

Circularity Check

0 steps flagged

No circularity: empirical reconstruction pipeline whose SOTA metrics are measured on held-out real cameras never used in training or diffusion conditioning.

full rationale

StudioRecon is a systems/method paper that assembles existing components (GEN3C video diffusion for background densification, monocular pose estimators + multi-view triangulation for SMPL initialization, 3DGS optimization, Difix3D+ single-step enhancement with RAFT-based EMA injection). There is no claimed first-principles derivation, uniqueness theorem, or analytic prediction that could reduce to its inputs by construction. All free parameters (affinity weights wp/w heta, loss lambdas, EMA rates, mask dilation, etc.) are ordinary hyperparameters; none is fitted to the evaluation metrics and then re-presented as a prediction. Quantitative claims (Tables 1–2, S4–S5) compare rendered novel views against real held-out cameras that are never seen by the pipeline or by GEN3C conditioning. Self-citations are absent from the load-bearing chain; the paper cites external models and datasets. Consequently the derivation chain is empty of circular steps and the evaluation is externally falsifiable.

Axiom & Free-Parameter Ledger

4 free parameters · 3 axioms · 2 invented entities

The central claim rests on the empirical success of a multi-stage engineering pipeline that imports several large pre-trained models and a handful of hand-chosen hyper-parameters; no new physical or mathematical axioms are postulated, but the method inherits the inductive biases of those models and the modeling choice that backgrounds and humans can be cleanly decoupled.

free parameters (4)
  • affinity weights w_p=0.9, w_theta=0.1 and distance threshold 0.3 m
    Hand-chosen combination of spatial and pose cues for cross-view identity association; ablation shows they matter but values are not derived.
  • loss weights (L1/SSIM/LPIPS/density/pose) and densification schedules for bg and human Gaussians
    Standard 3DGS hyper-parameters tuned for the sparse setting; affect final PSNR/LPIPS.
  • EMA injection strength, decay, K=3 previous frames, warp-error threshold tau_e=30
    Hand-tuned parameters of the motion-adaptive consistency injection that control temporal coherence vs. ghosting.
  • mask dilation 21 px, height oscillation +/-15 % for iterative refinement
    Ad-hoc geometric heuristics that keep humans out of background Gaussians and densify undersampled regions.
axioms (3)
  • domain assumption Camera-controlled video diffusion (GEN3C) can synthesize hundreds of novel views of a static background that are sufficiently consistent with the sparse real observations to serve as photometric supervision.
    Invoked throughout Sec. 3.1 and 3.3; the paper shows it works better than pure sparse 3DGS but does not prove geometric fidelity.
  • domain assumption SMPL parametric body model plus multi-view triangulation supplies a sufficiently accurate geometric prior for deformable human Gaussians even under heavy occlusion and low overlap.
    Core of Sec. 3.2; failure would collapse human reconstruction quality.
  • ad hoc to paper Backgrounds and humans can be optimized separately and later composited without irreversible lighting or contact artifacts that the recursive enhancer cannot fix.
    Central design choice of the whole pipeline (abstract and Sec. 3).
invented entities (2)
  • StudioRecon pipeline (sparse-to-dense view synthesis + geometry-driven multi-view SMPL + decoupled Gaussians + recursive enhancement with motion-adaptive consistency injection) no independent evidence
    purpose: End-to-end system that turns four low-overlap videos into editable 4D Gaussians and temporally coherent free-viewpoint video.
    The composite system is new; individual modules are assembled from prior work.
  • Motion-adaptive consistency injection (EMA of flow-warped previous enhanced frames weighted by per-pixel warp confidence) no independent evidence
    purpose: Convert a single-step image diffusion enhancer into a temporally stable video enhancer without multi-step sampling.
    Specific mechanism introduced in Sec. 3.4; ablation shows it reduces Warp-L2.

pith-pipeline@v1.1.0-grok45 · 25765 in / 2880 out tokens · 34505 ms · 2026-07-13T05:12:58.385913+00:00 · methodology

0 comments
read the original abstract

Existing volumetric capture of dynamic human performance achieves high fidelity with dense camera arrays. However, in real-world scenarios, only a handful of low-overlap cameras are available, which degrades the output quality and leaves large areas unobserved. Recent 4D reconstruction methods have focused on low-overlap settings, yet they still produce noticeable artifacts in under-observed regions. Video diffusion models have emerged as another option, but they show geometrically inconsistent results for humans. To address these limitations, we propose StudioRecon, a pipeline that reconstructs 4D human scenes from sparse, low-overlap cameras by decoupling background and humans. We densify background supervision by synthesizing hundreds of camera-controlled novel views with a video diffusion model. We also robustly initialize deformable Gaussian humans with cross-view identity association and triangulated multi-view keypoint fitting. Finally, our recursive enhancement module with motion-adaptive consistency injection harmonizes the composed output, thereby further avoiding remaining artifacts. We achieve state-of-the-art novel view synthesis across four real-world datasets and demonstrate applications such as novel trajectory rendering and human replacement.

Figures

Figures reproduced from arXiv: 2607.09125 by Daneul Kim, Jaesik Park, Minhyuk Hwang, Sangmin Kim, Seunguk Do.

Figure 1
Figure 1. Figure 1: Given only as few as four sparse, low-overlap input videos (left), StudioRecon first reconstructs decoupled Gaussians for background and humans (right). [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the proposed StudioRecon. Our pipeline consists of four stages: (1) sparse-to-dense view synthesis using camera-controlled video diffusion, [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of our recursive enhancement module (Sec. 3.4). [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Schematic of motion-adaptive consistency injection. For each frame, [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison on 360◦ scenes (Legoassemble, Grappling, Sword, Karate). Our method produces sharper backgrounds and more robust human reconstructions than baselines. where 𝑤𝑏 (𝝁𝑐 ) are skinning weights queried from a precomputed voxel grid derived from the SMPL model, and G𝑡 𝑏 ∈ 𝑆𝐸(3) are the bone transformation matrices for timestep 𝑡. To capture fine-grained appearance changes, we employ a tempor… view at source ↗
Figure 6
Figure 6. Figure 6: Additional qualitative comparison (Tennis, Fencing, Dance, Yoga) from EgoHumans, Mobile Stage, and SelfCap. Our method produces sharper reconstructions with better human-scene separation. Dance © Xu et al. (Mobile Stage); Yoga © Xu et al. (SelfCap), used with permission. w/o enhancement w/ enhancement (ours) [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Ablation on recursive enhancement (Sec. 3.4). Raw renders (left) contain blur and geometric instabilities. Enhancement (right) produces clean, [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Qualitative results on EgoExo-4D [Grauman et al. 2024]. Our method applies to diverse activities and environments. [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Effect of iterative refinement on background reconstruction (Sec. 3.3). [PITH_FULL_IMAGE:figures/full_fig_p010_9.png] view at source ↗
Figure 11
Figure 11. Figure 11: Visualization of motion-adaptive consistency injection (Sec. 3.4). From [PITH_FULL_IMAGE:figures/full_fig_p010_11.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

69 extracted references · 3 linked inside Pith

  1. [1]

    Wang, Ruicheng and Xu, Sicheng and Dai, Cassie and Xiang, Jianfeng and Deng, Yu and Tong, Xin and Yang, Jiaolong , booktitle=

  2. [2]

    2024 , organization=

    Chen, Yuedong and Xu, Haofei and Zheng, Chuanxia and Zhuang, Bohan and Pollefeys, Marc and Geiger, Andreas and Cham, Tat-Jen and Cai, Jianfei , booktitle=. 2024 , organization=

    2024

  3. [3]

    arXiv preprint arXiv:2511.00560 , year=

    4D Neural Voxel Splatting: Dynamic Scene Rendering with Voxelized Gaussian Splatting , author=. arXiv preprint arXiv:2511.00560 , year=

    Pith/arXiv arXiv

  4. [4]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Generative sparse-view Gaussian splatting , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  5. [5]

    2024 , organization=

    Liu, Xinhang and Chen, Jiaben and Kao, Shiu-hong and Tai, Yu-Wing and Tang, Chi-Keung , booktitle=. 2024 , organization=

    2024

  6. [6]

    Park, Hyunwoo and Ryu, Gun and Kim, Wonjun , booktitle=

  7. [7]

    ACM transactions on graphics (TOG) , volume=

    Free-viewpoint video of human actors , author=. ACM transactions on graphics (TOG) , volume=. 2003 , publisher=

    2003

  8. [8]

    2016 , publisher=

    Dou, Mingsong and Khamis, Sameh and Degtyarev, Yury and Davidson, Philip and Fanello, Sean Ryan and Kowdle, Adarsh and Escolano, Sergio Orts and Rhemann, Christoph and Kim, David and Taylor, Jonathan and others , journal=. 2016 , publisher=

    2016

  9. [9]

    ACM SIGGRAPH 2008 papers , pages=

    Performance capture from sparse multi-view video , author=. ACM SIGGRAPH 2008 papers , pages=

    2008

  10. [10]

    Newcombe, Richard A and Fox, Dieter and Seitz, Steven M , booktitle=

  11. [11]

    2021 , publisher=

    Mildenhall, Ben and Srinivasan, Pratul P and Tancik, Matthew and Barron, Jonathan T and Ramamoorthi, Ravi and Ng, Ren , journal=. 2021 , publisher=

    2021

  12. [12]

    ACM Trans

    3D Gaussian splatting for real-time radiance field rendering , author=. ACM Trans. Graph. , volume=

  13. [13]

    Li, Zhengqi and Wang, Qianqian and Cole, Forrester and Tucker, Richard and Snavely, Noah , booktitle=

  14. [14]

    Shape of motion:

    Wang, Qianqian and Ye, Vickie and Gao, Hang and Zeng, Weijia and Austin, Jake and Li, Zhengqi and Kanazawa, Angjoo , booktitle=. Shape of motion:

  15. [15]

    2025 , organization=

    Tan, Jeff and Xiang, Donglai and Tulsiani, Shubham and Ramanan, Deva and Yang, Gengshan , booktitle=. 2025 , organization=

    2025

  16. [16]

    Deformable

    Yang, Ziyi and Gao, Xinyu and Zhou, Wen and Jiao, Shaohui and Zhang, Yuqing and Jin, Xiaogang , booktitle=. Deformable

  17. [17]

    SIGGRAPH Asia 2024 Conference Papers , pages=

    Dynamic Gaussian marbles for novel view synthesis of casual monocular videos , author=. SIGGRAPH Asia 2024 Conference Papers , pages=

    2024

  18. [18]

    Lei, Jiahui and Weng, Yijia and Harley, Adam W and Guibas, Leonidas and Daniilidis, Kostas , booktitle=

  19. [19]

    Luiten, Jonathon and Kopanas, Georgios and Leibe, Bastian and Ramanan, Deva , booktitle=. Dynamic. 2024 , organization=

    2024

  20. [20]

    Wu, Guanjun and Yi, Taoran and Fang, Jiemin and Xie, Lingxi and Zhang, Xiaopeng and Wei, Wei and Liu, Wenyu and Tian, Qi and Wang, Xinggang , booktitle=

  21. [21]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Fridovich-Keil, Sara and Meanti, Giacomo and Warburg, Frederik Rahb. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  22. [22]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    MonoFusion: Sparse-View 4D Reconstruction via Monocular Fusion , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  23. [23]

    Kocabas, Muhammed and Chang, Jen-Hao Rick and Gabriel, James and Tuzel, Oncel and Ranjan, Anurag , booktitle=

  24. [24]

    ACM Transactions on Graphics (TOG) , volume=

    Echoes of the Coliseum: Towards 3D Live streaming of Sports Events , author=. ACM Transactions on Graphics (TOG) , volume=. 2025 , publisher=

    2025

  25. [25]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    ShowMak3r: Compositional TV Show Reconstruction , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  26. [26]

    HSR: holistic

    Xue, Lixin and Guo, Chen and Zheng, Chengwei and Wang, Fangjinghua and Jiang, Tianjian and Ho, Hsuan-I and Kaufmann, Manuel and Song, Jie and Hilliges, Otmar , booktitle=. HSR: holistic. 2024 , organization=

    2024

  27. [27]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    ODHSR: Online Dense 3D Reconstruction of Humans and Scenes from Monocular Videos , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  28. [28]

    Bai, Jianhong and Xia, Menghan and Fu, Xiao and Wang, Xintao and Mu, Lianrui and Cao, Jinwen and Liu, Zuozhu and Hu, Haoji and Bai, Xiang and Wan, Pengfei and others , journal=

  29. [29]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Ren, Xuanchi and Shen, Tianchang and Huang, Jiahui and Ling, Huan and Lu, Yifan and Nimier-David, Merlin and M. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  30. [30]

    SIGGRAPH Asia 2022 Conference Papers , pages=

    Fast dynamic radiance fields with time-aware neural voxels , author=. SIGGRAPH Asia 2022 Conference Papers , pages=

    2022

  31. [31]

    arXiv preprint arXiv:2511.05152 , year=

    Splatography: Sparse multi-view dynamic Gaussian Splatting for filmmaking challenges , author=. arXiv preprint arXiv:2511.05152 , year=

    Pith/arXiv arXiv

  32. [32]

    International Conference on Learning Representations (ICLR) , year=

    Pi^3 : Permutation-Equivariant Visual Geometry Learning , author=. International Conference on Learning Representations (ICLR) , year=

  33. [33]

    Carion, Nicolas and Gustafson, Laura and Hu, Yuan-Ting and Debnath, Shoubhik and Hu, Ronghang and Suris, Didac and Ryali, Chaitanya and Alwala, Kalyan Vasudev and Khedr, Haitham and Huang, Andrew and others , journal=

  34. [34]

    International Conference on Learning Representations (ICLR) , year=

    CoMotion: Concurrent Multi-person 3D Motion , author=. International Conference on Learning Representations (ICLR) , year=

  35. [35]

    Neural localizer fields for continuous

    S. Neural localizer fields for continuous. Advances in Neural Information Processing Systems , volume=

  36. [36]

    arXiv preprint arXiv:2510.24464 , year=

    Kineo: Calibration-Free Metric Motion Capture From Sparse RGB Cameras , author=. arXiv preprint arXiv:2510.24464 , year=

    arXiv

  37. [37]

    Hu, Shoukang and Hu, Tao and Liu, Ziwei , booktitle=

  38. [38]

    Seminal Graphics Papers: Pushing the Boundaries, Volume 2 , pages=

    SMPL: A skinned multi-person linear model , author=. Seminal Graphics Papers: Pushing the Boundaries, Volume 2 , pages=

  39. [39]

    Grauman, Kristen and Westbury, Andrew and Torresani, Lorenzo and Kitani, Kris and Malik, Jitendra and Afouras, Triantafyllos and Ashutosh, Kumar and Baiyya, Vijay and Bansal, Siddhant and Boote, Bikram and others , booktitle=

  40. [40]

    Wang, Yanhui and Bao, Jianmin and Weng, Wenming and Feng, Ruoyu and Yin, Dacheng and Yang, Tao and Zhang, Jingxu and Dai, Qi and Zhao, Zhiyuan and Wang, Chunyu and others , booktitle=

  41. [41]

    Khirodkar, Rawal and Bansal, Aayush and Ma, Lingni and Newcombe, Richard and Vo, Minh and Kitani, Kris , booktitle=

  42. [42]

    Khirodkar, Rawal and Song, Jyun-Ting and Cao, Jinkun and Luo, Zhengyi and Kitani, Kris , journal=

  43. [43]

    Understanding

    Nilsson, Jim and Akenine-M. Understanding. arXiv preprint arXiv:2006.13846 , year=

    Pith/arXiv arXiv 2006

  44. [44]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    The unreasonable effectiveness of deep features as a perceptual metric , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  45. [45]

    Qiu, Lingteng and Gu, Xiaodong and Li, Peihao and Zuo, Qi and Shen, Weichao and Zhang, Junfei and Qiu, Kejie and Yuan, Weihao and Chen, Guanying and Dong, Zilong and others , journal=

  46. [46]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Spacetime Gaussian feature splatting for real-time dynamic view synthesis , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  47. [47]

    Xu, Zhen and Peng, Sida and Lin, Haotong and He, Guangzhao and Sun, Jiaming and Shen, Yujun and Bao, Hujun and Zhou, Xiaowei , booktitle=

  48. [48]

    ACM Transactions on Graphics (TOG) , volume=

    Representing long volumetric video with temporal Gaussian hierarchy , author=. ACM Transactions on Graphics (TOG) , volume=. 2024 , publisher=

    2024

  49. [49]

    International Conference on Learning Representations (ICLR) , year=

    How I Warped Your Noise: a Temporally-Correlated Noise Prior for Diffusion Models , author=. International Conference on Learning Representations (ICLR) , year=

  50. [50]

    Advances in Neural Information Processing Systems , volume=

    Warped diffusion: Solving video inverse problems with image diffusion models , author=. Advances in Neural Information Processing Systems , volume=

  51. [51]

    Zhou, Shangchen and Yang, Peiqing and Wang, Jianyi and Luo, Yihang and Loy, Chen Change , booktitle=

  52. [52]

    European conference on computer vision , pages=

    Motion-guided latent diffusion for temporally consistent real-world video super-resolution , author=. European conference on computer vision , pages=. 2024 , organization=

    2024

  53. [53]

    Yeh, Chang-Han and Shiu, Hau-Shiang and Lin, Chin-Yang and Wang, Zhixiang and Hsiao, Chi-Wei and Chen, Ting-Hsuan and Liu, Yu-Lun , journal=

  54. [54]

    Liu, Xi and Zhou, Chaoyi and Huang, Siyu , journal=

  55. [55]

    Wu, Jay Zhangjie and Zhang, Yuxuan and Turki, Haithem and Ren, Xuanchi and Gao, Jun and Shou, Mike Zheng and Fidler, Sanja and Gojcic, Zan and Ling, Huan , booktitle=

  56. [56]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Effective whole-body pose estimation with two-stages distillation , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  57. [57]

    European Conference on Computer Vision , pages=

    Spherical linear interpolation and text-anchoring for zero-shot composed image retrieval , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    2024

  58. [58]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    FreeTimeGS: Free Gaussian Primitives at Anytime Anywhere for Dynamic Scene Reconstruction , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  59. [59]

    International Conference on Learning Representations (ICLR) , year =

    Real-time Photorealistic Dynamic Scene Representation and Rendering with 4D Gaussian Splatting , author =. International Conference on Learning Representations (ICLR) , year =

  60. [60]

    Li, Jiahe and Zhang, Jiawei and Bai, Xiao and Zheng, Jin and Ning, Xin and Zhou, Jun and Gu, Lin , booktitle=

  61. [61]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Structure-from-motion revisited , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  62. [62]

    Jin, Yudong and Peng, Sida and Wang, Xuan and Xie, Tao and Xu, Zhen and Yang, Yifan and Shen, Yujun and Bao, Hujun and Zhou, Xiaowei , booktitle=

  63. [63]

    360-degree human video generation with

    Shao, Ruizhi and Pang, Youxin and Zheng, Zerong and Sun, Jingxiang and Liu, Yebin , journal=. 360-degree human video generation with. 2024 , publisher=

    2024

  64. [64]

    Wu, Rundi and Gao, Ruiqi and Poole, Ben and Trevithick, Alex and Zheng, Changxi and Barron, Jonathan T and Holynski, Aleksander , booktitle=

  65. [65]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Towards Explicit Exoskeleton for the Reconstruction of Complicated 3D Human Avatars , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  66. [66]

    Expressive whole-body

    Moon, Gyeongsik and Shiratori, Takaaki and Saito, Shunsuke , booktitle=. Expressive whole-body. 2024 , organization=

    2024

  67. [67]

    Huang, Yi-Hua and Sun, Yang-Tian and Yang, Ziyi and Lyu, Xiaoyang and Cao, Yan-Pei and Qi, Xiaojuan , booktitle=

  68. [68]

    Duan, Yuanxing and Wei, Fangyin and Dai, Qiyu and He, Yuhang and Chen, Wenzheng and Chen, Baoquan , booktitle=

  69. [69]

    2020 , organization=

    Teed, Zachary and Deng, Jia , booktitle=. 2020 , organization=

    2020