Pith. sign in

REVIEW 3 major objections 6 minor 156 references

Complete animatable 3D people can be built from ordinary YouTube clips even when most of the body is hidden.

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 22:47 UTC pith:XPSH7ZTP

load-bearing objection Solid systems paper that makes occluded YouTube video usable for animatable 3DGS avatars; the real novelty is the pipeline glue, and the load-bearing diffusion hallucination is only partly stress-tested. the 3 major comments →

arxiv 2603.17975 v2 pith:XPSH7ZTP submitted 2026-03-18 cs.CV

AHOY! Animatable Humans under Occlusion from YouTube Videos with Gaussian Splatting and Video Diffusion Priors

Aymen Mir , Riza Alp Guler , Xiangjun Tang , Peter Wonka , Gerard Pons-Moll This is my paper
classification cs.CV
keywords 3D Gaussian splattinganimatable avatarsocclusionvideo diffusionmonocular reconstructionpose-dependent appearanceYouTube humans
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.

Most real video of people is messy: furniture, other people, or the camera framing hide large parts of the body, so existing avatar methods that need a fully visible subject cannot use it. AHOY claims that identity-tuned video diffusion can fill those gaps well enough to act as multi-view training data for a 3D Gaussian avatar. The pipeline first builds a coarse avatar from the visible pixels, renders it in planned turn and pose sequences, then passes those renders through an identity-finetuned diffusion model so the missing surfaces appear with the right clothing and identity. Those hallucinated videos supervise a second-stage avatar that learns how appearance changes with pose. A split between the pose that drives the appearance map and the pose that deforms the Gaussians absorbs the geometric noise that diffusion introduces, and a separate head path keeps the face recognizable. The result is a complete, reposable person that can be dropped into phone-captured 3D scenes.

Core claim

A complete, animatable 3D Gaussian avatar of a specific person can be reconstructed from heavily occluded monocular video by treating identity-finetuned video diffusion (applied via inversion to coarse avatar renders) as dense multi-view supervision, then training pose-dependent Gaussian maps with a map-pose versus LBS-pose split and separate head supervision.

What carries the argument

Hallucination-as-supervision via RF-Inversion of coarse avatar renders through an identity-finetuned video diffusion model, combined with a two-stage upgrade from canonical to pose-dependent Gaussian maps and a map-pose/LBS-pose decoupling that absorbs multi-view inconsistency.

Load-bearing premise

That the diffusion model, once identity-tuned, invents never-seen body surfaces accurately enough to serve as geometric and photometric training signal rather than merely plausible fiction.

What would settle it

On multi-view captures with known side views of body regions that were fully occluded in the monocular input, measure whether the reconstructed avatar matches those held-out ground-truth views; systematic mismatch on never-observed surfaces would falsify the claim that the hallucinated supervision is accurate enough.

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

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

3 major / 6 minor

Summary. AHOY reconstructs complete, animatable 3D Gaussian avatars from monocular in-the-wild video under heavy occlusion. From partial observations it builds a coarse canonical 3DGS avatar (DensePose UV atlas, FLUX inpainting, multi-view diffusion, StyleUNet + LBS; Sec. 3.1, Eq. 1). It then identity-finetunes Wan 2.2 with LoRA on the input clip, renders the coarse avatar under structured turn/static motions, and refines those renders via RF-Inversion to produce hallucinated multi-view supervision videos (Sec. 3.2, Eqs. 2–4). A second stage upgrades to pose-dependent Gaussian maps, with map-pose vs. LBS-pose decoupling and joint pose/camera correction to absorb multi-view inconsistency of the generated data, plus a separate FLAME head path for identity (Sec. 3.3, Eqs. 5–8). Evaluation on occluded YouTube clips and BEHAVE (novel view/pose) reports large gains over single-image baselines (LHM, IDOL, PSHuman, SyncHuman; Tabs. 1–2), with ablations for coarse-only, no RF-Inversion, no map/LBS decoupling, and no head split (Tab. 3). Avatars are further animated into cell-phone 3DGS scenes.

Significance. If the result holds, the paper substantially expands the usable data regime for photorealistic human avatars: from controlled or fully visible monocular capture to everyday occluded YouTube-style footage. The engineering combination—identity-finetuned video diffusion as dense supervision, canonical-to-pose-dependent Gaussian maps, map/LBS decoupling, and head/body split—is a coherent response to the two real obstacles (never-observed surfaces and lack of multi-view per pose). Quantitative tables on both monocular YouTube and multi-view BEHAVE, plus component ablations that move metrics in the expected direction, give the claim more empirical weight than typical single-image avatar papers. The scene-composition demos further show practical utility. The main open scientific question is how much of the unobserved appearance is true recovery versus plausible invention; the paper itself flags this limitation.

major comments (3)
  1. Sec. 3.2.3 and conclusion: the load-bearing claim is that RF-Inversion through the identity-finetuned model yields appearance for never-observed body regions that is accurate enough to supervise 3DGS, not only plausible invention. BEHAVE novel-view/novel-pose gains (Tab. 1: 24.12 / 22.81 PSNR) are the strongest external evidence, but the protocol uses only the occluded front Kinect as input; side cameras can still observe surfaces that were never in the front stream. Ablation (B) (Tab. 3) removes RF-Inversion and drops metrics, yet does not isolate fully invented vs. partially observed surfaces. A targeted analysis—e.g., masking regions never visible from the front across the sequence and reporting side-view error only on those pixels, or qualitative multi-view comparisons of back/inner-arm regions against BEHAVE GT—would make the central claim falsifiable rather than only directionally
  2. Sec. 4.1–4.4 and Tabs. 1–2: all animatable baselines (LHM, IDOL) and static multi-view diffusion methods (PSHuman, SyncHuman) are single-image methods. Even when given the authors’ unoccluded canonical-pose image, they do not exploit the full monocular video or multi-frame occlusion reasoning that AHOY uses. The large gaps are therefore partly expected from input modality. The paper notes that several closer video-based or multi-view diffusion avatar methods lack public code; without at least one stronger video-aware baseline (or a controlled multi-frame adaptation of an existing method), the SOTA claim overstates the comparison. Clarifying this limitation and, if possible, adding one video-capable baseline would strengthen the experimental case.
  3. Sec. 3.3.1–3.3.2: map-pose/LBS-pose decoupling is presented as absorbing multi-view inconsistency of independently sampled diffusion videos. The free per-frame corrections {Δθ_n,t, Δγ_n,t} can also absorb residual pose error and camera drift, so it is unclear how much of the gain is geometric consistency vs. extra degrees of freedom. Ablation (C) (Tab. 3) removes the decoupling and hurts metrics, but does not report the magnitude of the optimized corrections or a control that optimizes the same corrections without pose-dependent maps. A short analysis of correction norms and a control experiment would pin down whether the architectural insight is necessary or whether extra pose/camera freedom alone explains much of the improvement.
minor comments (6)
  1. Eq. (1) and Sec. 3.1: the coarse loss uses L1 + perceptual loss with occlusion masks; the perceptual term is only mentioned in text. Stating the full coarse objective (including λ_per) in the main text would match the care given to Eq. (8).
  2. Sec. 3.2.2: the design of structured motions (which actions, how many turns vs. static multi-angle sequences, how synchronization is done for sitting) is described at a high level. A short table or supplementary list of the N sequences would aid reproducibility.
  3. Fig. 3–6 captions and Tab. 1: “Zoom In” and qualitative claims would be easier to verify if the paper indicated which body regions were never observed in the input (e.g., back, occluded torso) rather than only showing overall frames.
  4. Related work (Sec. 2): several concurrent monocular/video avatar and diffusion-supervision works are cited; a clearer sentence distinguishing AHOY’s occlusion-first setting from sparse-view 3DGS-with-diffusion methods (e.g., those using video diffusion under sparse multi-view) would help positioning.
  5. Sec. 3.4 / inference: PCA pose projection is deferred to the supplementary; a one-sentence statement of the retained variance or projection rule in the main text would clarify the novel-pose protocol used in Tab. 1–2.
  6. Typos / polish: “everybodytype” (p. 2), occasional missing spaces around citations, and “in-effect 360°” hyphenation are minor but should be cleaned for camera-ready.

Circularity Check

0 steps flagged

No by-construction circularity: empirical pipeline with external multi-view benchmarks; generative prior use and self-citations are non-load-bearing.

full rationale

AHOY is an engineering reconstruction pipeline, not a first-principles derivation claiming to predict quantities from axioms. Identity LoRA finetuning (Eq. 2) plus RF-Inversion (Eq. 4) produces hallucinated supervision videos that are then used to train pose-dependent Gaussian maps; this is standard use of a generative prior, not a self-definitional reduction (the avatar parameters are optimized against the generated images plus observed frames, not algebraically identical to the LoRA weights). Quantitative claims rest on independent BEHAVE multi-view ground truth (held-out side cameras for novel-view and novel-pose PSNR/SSIM/LPIPS) and ablations that remove components and measure drops; these metrics are not forced by the training inputs. Self-citations (Animatable Gaussians/Gaspacho for map design, AHA! for scene composition) supply architectural building blocks, not uniqueness theorems or load-bearing premises that forbid alternatives. No equation equates a fitted parameter to a claimed prediction of the same quantity, and the paper itself flags the hallucination risk as a limitation rather than concealing it. The mild self-referential flavor of same-video held-out evaluation is ordinary for monocular reconstruction and is offset by the external BEHAVE protocol; it does not meet the threshold for circular steps under the stated patterns.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 3 invented entities

The claim rests on standard human body models and rendering (SMPL/LBS, 3DGS, DensePose UV), on pretrained generative models treated as identity/appearance oracles after LoRA finetuning, and on several paper-specific modeling choices (canonical-to-pose-dependent maps, map vs LBS pose split, head/body split, structured motion design, loss weights). Free parameters are mainly loss weights and optimization corrections; invented entities are architectural devices rather than new physics particles, with independent evidence only via ablations and external BEHAVE views.

free parameters (4)
  • Loss weights λ_body, λ_obs, λ_head, λ_per, λ_reg
    Balance body hallucination, observed frames, head path, perceptual, and skinning regularization in Eq. 8; values not reported in the main text but control the full-avatar objective.
  • Per-frame LBS pose corrections Δθ_n,t and camera corrections Δγ_n,t
    Optimized jointly with StyleUNet on hallucinated frames (Sec. 3.3.2) to absorb multi-view inconsistency; free degrees of freedom fitted per frame of generated data.
  • LoRA update AB and learnable identity token [v]
    Fitted on the unordered input video frames via flow-matching loss (Eq. 2) to bind subject identity in Wan 2.2; identity prior is data-fit, not derived.
  • Number and design of structured motion sequences N (turns vs static multi-angle)
    Hand-curated motion set that defines which poses get multi-view supervision; choice directly enables pose-dependent maps.
axioms (5)
  • domain assumption SMPL/LBS and DensePose UV provide a sufficiently accurate body topology and skinning for monocular occluded video after NLF pose estimation.
    Invoked throughout Sec. 3.1–3.3 for texture atlas, Gaussian templates, and deformation; monocular pose under occlusion is known to be noisy.
  • ad hoc to paper An identity-finetuned rectified-flow video model plus RF-Inversion preserves coarse pose/silhouette while replacing unobserved appearance with identity-consistent texture usable as 3D supervision.
    Core of Sec. 3.2.3; not a standard theorem—empirical assumption that diffusion fill-in is good enough for 3DGS training.
  • ad hoc to paper Sharing a single map pose across a turn (or synchronized static multi-view frames) yields valid multi-view supervision of one body configuration despite residual pose change during rotation.
    Sec. 3.2.2 and 3.3.1 treat turn frames as same-pose multi-view; approximate constancy of body action is assumed.
  • domain assumption Pretrained multi-view face diffusion (CAP4D) and FLUX inpainting supply identity-preserving head and missing-region appearance independent of Wan body videos.
    Sec. 3.1.1 and 3.3.3; relies on external generative models behaving as claimed.
  • domain assumption 3D Gaussian splatting with StyleUNet-predicted pose-dependent maps can represent high-fidelity clothed humans under LBS.
    Follows Animatable Gaussians / related 3DGS avatar literature cited in Sec. 2–3.
invented entities (3)
  • Hallucination-as-supervision pipeline (coarse render → RF-Inversion through identity LoRA → dense body videos) no independent evidence
    purpose: Create multi-view, multi-pose photometric supervision for body regions never observed in the input monocular video.
    Central methodological construct of the paper; evidence is ablations and BEHAVE gains, not an external physical measurement.
  • Map-pose / LBS-pose decoupling no independent evidence
    purpose: Keep StyleUNet Gaussian attributes consistent per body configuration while letting per-frame deformation absorb diffusion multi-view inconsistency.
    Introduced in Sec. 3.3.1–3.3.2; ablation (C) supports utility but the split is a modeling device without external independent existence.
  • Head/body split supervision via SMPL–FLAME partition no independent evidence
    purpose: Avoid identity drift from unreliable Wan faces by supervising head Gaussians with CAP4D multi-view faces only.
    Sec. 3.3.3; architectural partition justified by known diffusion face unreliability and ablation (D).

pith-pipeline@v1.1.0-grok45 · 26386 in / 4168 out tokens · 39450 ms · 2026-07-13T22:47:50.586219+00:00 · methodology

0 comments
read the original abstract

We present AHOY, a method for reconstructing complete, animatable 3D Gaussian avatars from in-the-wild monocular video despite heavy occlusion. Existing methods assume unoccluded input-a fully visible subject, often in a canonical pose-excluding the vast majority of real-world footage where people are routinely occluded by furniture, objects, or other people. Reconstructing from such footage poses fundamental challenges: large body regions may never be observed, and multi-view supervision per pose is unavailable. We address these challenges with four contributions: (i) a hallucination-as-supervision pipeline that uses identity-finetuned diffusion models to generate dense supervision for previously unobserved body regions; (ii) a two-stage canonical-to-pose-dependent architecture that bootstraps from sparse observations to full pose-dependent Gaussian maps; (iii) a map-pose/LBS-pose decoupling that absorbs multi-view inconsistencies from the generated data; (iv) a head/body split supervision strategy that preserves facial identity. We evaluate on YouTube videos and on multi-view capture data with significant occlusion and demonstrate state-of-the-art reconstruction quality. We also demonstrate that the resulting avatars are robust enough to be animated with novel poses and composited into 3DGS scenes captured using cell-phone video. Our project page is available at https://miraymen.github.io/ahoy/

Figures

Figures reproduced from arXiv: 2603.17975 by Aymen Mir, Gerard Pons-Moll, Peter Wonka, Riza Alp Guler, Xiangjun Tang.

Figure 1
Figure 1. Figure 1: Given a monocular YouTube video with occlusion, AHOY reconstructs a complete, animatable 3D human avatar using video diffusion priors and 3D Gaussian Splatting, enabling photorealistic human animation within 3D scenes. Abstract. We present AHOY, a method for reconstructing complete, an￾imatable 3D Gaussian avatars from in-the-wild monocular video despite heavy occlusion. Existing methods assume unoccluded … view at source ↗
Figure 2
Figure 2. Figure 2: Method overview. Block 1 (Sec. 3.1): We map observed textures from partially occluded video onto a canonical pose via DensePose UV correspondences, inpaint missing regions with FLUX, and generate multi-view canonical-pose images with multi-view diffusion to supervise a coarse 3DGS avatar using canonical Gaussian maps. Block 2 (Sec. 3.2): We finetune a video diffusion model (Wan 2.2) with LoRA to capture th… view at source ↗
Figure 3
Figure 3. Figure 3: Animation comparison (Zoom In). Top: canonical-pose input. Bottom (below line): occluded input. Ours produces higher-fidelity avatars in both settings. 3.3.4 Total objective. The full avatar is trained with the combined loss: L = λbodyLbody + λobsLobs + λheadLhead + λperLper + λregLreg, (8) where Lper is a perceptual loss [145] and Lreg regularizes the skinning weights toward SMPL-derived values. The weigh… view at source ↗
Figure 4
Figure 4. Figure 4: Static reconstruction comparison (Zoom In). [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Reconstruction under difficult fully visible poses. [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Novel-view evaluation on BEHAVE. We render each reconstruction from held-out BEHAVE side cameras and compare against the ground-truth side views [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Visual ablations. Each pair shows the ablated result (w/o) and the full pipeline. (A) Without hallucination, the coarse-only avatar lacks detail in unobserved regions. (B) Without RF-Inversion, hallucinated supervision retains rendering artifacts. (C) Without map/LBS decoupling, multi-view inconsistencies cause blurring. (D) With￾out the head/body split, facial identity is lost [PITH_FULL_IMAGE:figures/fu… view at source ↗
Figure 8
Figure 8. Figure 8: Animation in 3DGS scenes. Avatars from occluded YouTube video, animated and composited into 3DGS scenes from casual cell-phone video. for StyleUNet training and do not optimize LBS deformation, removing the decoupling that absorbs multi-view inconsistencies (Sec. 3.3.2). (D) Without head/body split. We supervise the entire avatar, including the head, using the hallucinated Wan videos, without the separate … 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

156 extracted references · 14 linked inside Pith

  1. [1]

    In: ACM SIGGRAPH Asia Conference Proceedings (2025)

    Abdal, R., Patashnik, O., Deyneka, E., Chen, H., Siarohin, A., Tulyakov, S., Cohen-Or, D., Aberman, K.: Zero-shot dynamic concept personalization with grid-based LoRA. In: ACM SIGGRAPH Asia Conference Proceedings (2025)

    2025

  2. [2]

    In: ACM SIGGRAPH Conference Proceedings (2025)

    Abdal, R., Patashnik, O., Skorokhodov, I., Menapace, W., Siarohin, A., Tulyakov, S., Cohen-Or, D., Aberman, K.: Dynamic concepts personalization from single videos. In: ACM SIGGRAPH Conference Proceedings (2025)

    2025

  3. [3]

    Abdal, R., Qin, Y., Wonka, P.: Image2stylegan: How to embed images into the stylegan latent space? In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2019)

    2019

  4. [4]

    Abdal, R., Qin, Y., Wonka, P.: Image2stylegan++: How to edit the embedded images? In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2020)

    2020

  5. [5]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    Abdal, R., Yifan, W., Shi, Z., Xu, Y., Po, R., Kuang, Z., Chen, Q., Yeung, D.Y., Wetzstein, G.: Gaussian shell maps for efficient 3d human generation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    2024

  6. [6]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2019)

    Alldieck, T., Magnor, M., Bhatnagar, B.L., Theobalt, C., Pons-Moll, G.: Learning to reconstruct people in clothing from a single RGB camera. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2019)

    2019

  7. [7]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2018)

    Alldieck, T., Magnor, M., Xu, W., Theobalt, C., Pons-Moll, G.: Video based reconstruction of 3d people models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2018)

    2018

  8. [8]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2021)

    Alldieck, T., Xu, H., Sminchisescu, C.: imGHUM: Implicit generative models of 3d human shape and articulated pose. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2021)

    2021

  9. [9]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2022)

    Barron,J.T.,Mildenhall,B.,Verbin,D.,Srinivasan,P.P.,Hedman,P.:Mip-nerf360: Unbounded anti-aliased neural radiance fields. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2022)

    2022

  10. [10]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2023) 16 A

    Barron, J.T., Mildenhall, B., Verbin, D., Srinivasan, P.P., Hedman, P.: Zip-nerf: Anti-aliased grid-based neural radiance fields. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2023) 16 A. Mir et al

    2023

  11. [11]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2022)

    Bhatnagar, B.L., Xie, X., Petrov, I.A., Sminchisescu, C., Theobalt, C., Pons- Moll, G.: Behave: Dataset and method for tracking human object interactions. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2022)

    2022

  12. [12]

    arXiv preprint arXiv:2311.15127 (2023)

    Blattmann, A., Dockhorn, T., Kulal, S., Mendelevitch, D., Kilian, M., Lorber, D., Levi, Y., English, Z., Voleti, V., Letts, A., Jampani, V., Rombach, R.: Stable video diffusion: Scaling latent video diffusion models to large datasets. arXiv preprint arXiv:2311.15127 (2023)

    Pith/arXiv arXiv 2023

  13. [13]

    In: European Conference on Computer Vision (ECCV) (2016)

    Bogo, F., Kanazawa, A., Lassner, C., Gehler, P., Romero, J., Black, M.J.: Keep it SMPL: Automatic estimation of 3D human pose and shape from a single image. In: European Conference on Computer Vision (ECCV) (2016)

    2016

  14. [14]

    Brooks, T., Peebles, B., Holmes, C., DePue, W., Guo, Y., Jing, L., Schnurr, D., Taylor, J., Luhman, T., Luhman, E., Ng, C., Wang, R., Ramesh, A.: Video generation models as world simulators.https://openai.com/research/video- generation-models-as-world-simulators(2024)

    2024

  15. [15]

    In: International Conference on 3D Vision (3DV) (2026)

    Buehler, M.C., Yuan, Y., Li, X., Huang, Y., Nagano, K., Iqbal, U.: Dream, lift, animate: From single images to animatable gaussian avatars. In: International Conference on 3D Vision (3DV) (2026)

    2026

  16. [16]

    arXiv preprint arXiv:2506.09113 (2025)

    ByteDance Seed Team: Seedance 1.0: Exploring the boundaries of video generation models. arXiv preprint arXiv:2506.09113 (2025)

    Pith/arXiv arXiv 2025

  17. [17]

    arXiv preprint arXiv:2512.13507 (2025)

    ByteDance Seed Team: Seedance 1.5 pro: A native audio-visual joint generation foundation model. arXiv preprint arXiv:2512.13507 (2025)

    Pith/arXiv arXiv 2025

  18. [18]

    https : / / seed

    ByteDance Seed Team: Seedance 2.0. https : / / seed . bytedance . com / en / seedance2_0(2026)

    2026

  19. [19]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    Cao, Y., Cao, Y.P., Han, K., Shan, Y., Wong, K.Y.K.: DreamAvatar: Text-and- shape guided 3D human avatar generation via diffusion models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    2024

  20. [20]

    arXiv preprint arXiv:2511.16719 (2025)

    Carion, N., et al.: Sam 3: Segment anything with concepts. arXiv preprint arXiv:2511.16719 (2025)

    Pith/arXiv arXiv 2025

  21. [21]

    In: International Conference on Machine Learning (ICML) (2024)

    Chang, D., Shi, Y., Gao, Q., Xu, H., Fu, J., Song, G., Yan, Q., Zhu, Y., Yang, X., Soleymani, M.: MagicPose: Realistic human poses and facial expressions retargeting with identity-aware diffusion. In: International Conference on Machine Learning (ICML) (2024)

    2024

  22. [22]

    In: Advances in Neural Information Processing Systems (NeurIPS) (2025)

    Chen, W., Li, P., Zheng, W., Zhao, C., Li, M., Zhu, Y., Dou, Z., Wang, R., Liu, Y.: Synchuman: Synchronizing 2D and 3D generative models for single-view human reconstruction. In: Advances in Neural Information Processing Systems (NeurIPS) (2025)

    2025

  23. [23]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2021)

    Chen, X., Zheng, Y., Black, M.J., Hilliges, O., Geiger, A.: Snarf: Differentiable forward skinning for animating non-rigid neural implicit shapes. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2021)

    2021

  24. [24]

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

    Deng, B., Lewis, J.P., Jeruzalski, T., Pons-Moll, G., Hinton, G., Norouzi, M., Tagliasacchi, A.: Nasa neural articulated shape approximation. In: European Conference on Computer Vision (ECCV) (2020)

    2020

  25. [25]

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

    Dhamo, H., Nie, Y., Moreau, A., Song, J., Shaw, R., Zhou, Y., Pérez-Pellitero, E.: Headgas: Real-time animatable head avatars via 3d gaussian splatting. In: European Conference on Computer Vision (ECCV) (2024)

    2024

  26. [26]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025) AHOY: Animatable Humans under Occlusion from YouTube Videos 17

    Dong, Z., Duan, L., Song, J., Black, M.J., Geiger, A.: MoGA: 3D generative avatar prior for monocular gaussian avatar reconstruction. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025) AHOY: Animatable Humans under Occlusion from YouTube Videos 17

    2025

  27. [27]

    In: Advances in Neural Information Processing Systems (NeurIPS) (2024)

    Dravid, A., Gandelsman, Y., Wang, K.C., Abdal, R., Wetzstein, G., Efros, A.A., Aberman, K.: Interpreting the weight space of customized diffusion models. In: Advances in Neural Information Processing Systems (NeurIPS) (2024)

    2024

  28. [28]

    In: International Conference on Machine Learning (ICML) (2024)

    Esser, P., Kulal, S., Blattmann, A., Entezari, R., Müller, J., Saini, H., Levi, Y., Lorenz, D., Sauer, A., Boesel, F., Podell, D., Dockhorn, T., English, Z., Rombach, R.: Scaling rectified flow transformers for high-resolution image synthesis. In: International Conference on Machine Learning (ICML) (2024)

    2024

  29. [29]

    arXiv preprint arXiv:2502.04847 (2025)

    Gan, Q., Ren, Y., Zhang, C., Ye, Z., Xie, P., Yin, X., Yuan, Z., Peng, B., Zhu, J.: HumanDiT: Pose-guided diffusion transformer for long-form human motion video generation. arXiv preprint arXiv:2502.04847 (2025)

    Pith/arXiv arXiv 2025

  30. [30]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    Guédon, A., Lepetit, V.: Sugar: Surface-aligned gaussian splatting for efficient 3d mesh reconstruction and high-quality mesh rendering. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    2024

  31. [31]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2018)

    Güler, R.A., Neverova, N., Kokkinos, I.: DensePose: Dense human pose estimation in the wild. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2018)

    2018

  32. [32]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2023)

    Guo, C., Jiang, T., Chen, X., Song, J., Hilliges, O.: Vid2avatar: 3d avatar re- construction from videos in the wild via self-supervised scene decomposition. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2023)

    2023

  33. [33]

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

    Guo, C., Li, J., Kant, Y., Sheikh, Y., Saito, S., Cao, C.: Vid2avatar-pro: Authentic avatar from videos in the wild via universal prior. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2025)

    2025

  34. [34]

    arXiv preprint arXiv:2407.03168 (2024)

    Guo, J., Zhang, D., Liu, X., Zhong, Z., Zhang, Y., Wan, P., Zhang, D.: LivePortrait: Efficient portrait animation with stitching and retargeting control. arXiv preprint arXiv:2407.03168 (2024)

    Pith/arXiv arXiv 2024

  35. [35]

    In: International Conference on Learning Representations (ICLR) (2024)

    Guo, Y., Yang, C., Rao, A., Liang, Z., Wang, Y., Qiao, Y., Agrawala, M., Lin, D., Dai, B.: AnimateDiff: Animate your personalized text-to-image diffusion models without specific tuning. In: International Conference on Learning Representations (ICLR) (2024)

    2024

  36. [36]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2021)

    Guzov, V., Mir, A., Sattler, T., Pons-Moll, G.: Human poseitioning system (hps): 3d human pose estimation and self-localization in large scenes from body-mounted sensors. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2021)

    2021

  37. [37]

    Proceedings of the ACM on Computer Graphics and Interactive Techniques6(3) (2023)

    Habermann, M., Liu, L., Xu, W., Pons-Moll, G., Zollhoefer, M., Theobalt, C.: Hdhumans: A hybrid approach for high-fidelity digital humans. Proceedings of the ACM on Computer Graphics and Interactive Techniques6(3) (2023)

    2023

  38. [38]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2021)

    He, T., Xu, Y., Saito, S., Soatto, S., Tung, T.: Arch++: Animation-ready clothed human reconstruction revisited. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2021)

    2021

  39. [39]

    arXiv preprint arXiv:2408.14211 (2024)

    He, X., Li, X., Kang, D., Ye, J., Zhang, C., Chen, L., Gao, X., Zhang, H., Wu, Z., Zhuang, H.: MagicMan: Generative novel view synthesis of humans with 3D-aware diffusion and iterative refinement. arXiv preprint arXiv:2408.14211 (2024)

    Pith/arXiv arXiv 2024

  40. [40]

    In: Advances in Neural Information Processing Systems (NeurIPS) (2020)

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. In: Advances in Neural Information Processing Systems (NeurIPS) (2020)

    2020

  41. [41]

    In: International Conference on Learning Representations (ICLR) (2024) 18 A

    Hong, Y., Zhang, K., Gu, J., Bi, S., Zhou, Y., Liu, D., Liu, F., Sunkavalli, K., Bui, T., Tan, H.: LRM: Large reconstruction model for single image to 3D. In: International Conference on Learning Representations (ICLR) (2024) 18 A. Mir et al

    2024

  42. [42]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    Hu, L., Gao, X., Zhang, P., Sun, K., Zhang, B., Bo, L.: Animate anyone: Consistent and controllable image-to-video synthesis for character animation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    2024

  43. [43]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    Hu, L., Wang, G., Shen, Z., Gao, X., Meng, D., Zhuo, L., Zhang, P., Zhang, B., Bo, L.: Animate anyone 2: High-fidelity character image animation with environ- ment affordance. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    2025

  44. [44]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    Hu, L., Zhang, H., Zhang, Y., Zhou, B., Liu, B., Zhang, S., Nie, L.: Gaussianavatar: Towards realistic human avatar modeling from a single video via animatable 3d gaussians. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    2024

  45. [45]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    Hu, S., Hu, T., Liu, Z.: GauHuman: Articulated gaussian splatting from monocular human videos. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    2024

  46. [46]

    In: ACM SIGGRAPH Conference Proceedings (2024)

    Huang, B., Yu, Z., Chen, A., Geiger, A., Gao, S.: 2d gaussian splatting for geo- metrically accurate radiance fields. In: ACM SIGGRAPH Conference Proceedings (2024)

    2024

  47. [47]

    In: International Conference on 3D Vision (3DV) (2024)

    Huang, Y., Yi, H., Xiu, Y., Liao, T., Tang, J., Cai, D., Thies, J.: TeCH: Text- guided reconstruction of lifelike clothed humans. In: International Conference on 3D Vision (3DV) (2024)

    2024

  48. [48]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    Huang, Y., Yuan, Y., Li, X., Kautz, J., Iqbal, U.: Adahuman: Animatable detailed 3D human generation with compositional multiview diffusion. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    2025

  49. [49]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2020)

    Huang, Z., Xu, Y., Lassner, C., Li, H., Tung, T.: Arch: Animatable reconstruction of clothed humans. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2020)

    2020

  50. [50]

    arXiv preprint arXiv:2408.07009 (2024)

    Imagen Team Google: Imagen 3. arXiv preprint arXiv:2408.07009 (2024)

    arXiv 2024

  51. [51]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2022)

    Jiang, B., Hong, Y., Bao, H., Zhang, J.: SelfRecon: Self reconstruction your digital avatar from monocular video. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2022)

    2022

  52. [52]

    In: European Conference on Computer Vision (ECCV) (2022)

    Jiang, W., Yi, K.M., Samei, G., Tuzel, O., Ranjan, A.: Neuman: Neural human radiance field from a single video. In: European Conference on Computer Vision (ECCV) (2022)

    2022

  53. [53]

    ACM Transactions on Graphics43(6) (2024)

    Jiang, Y., Shen, Z., Hong, Y., Guo, C., Wu, Y., Zhang, Y., Yu, J., Xu, L.: Robust dual gaussian splatting for immersive human-centric volumetric videos. ACM Transactions on Graphics43(6) (2024)

    2024

  54. [54]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    Jiang, Z., Han, Z., Mao, C., Zhang, J., Pan, Y., Liu, Y.: VACE: All-in-one video creation and editing. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    2025

  55. [55]

    In: ACM SIGGRAPH Conference Proceedings (2025)

    Junkawitsch, H., Sun, G., Zhu, H., Theobalt, C., Habermann, M.: Eva: Expres- sive virtual avatars from multi-view videos. In: ACM SIGGRAPH Conference Proceedings (2025)

    2025

  56. [56]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2018)

    Kanazawa, A., Black, M.J., Jacobs, D.W., Malik, J.: End-to-end recovery of human shape and pose. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2018)

    2018

  57. [57]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024) AHOY: Animatable Humans under Occlusion from YouTube Videos 19

    Keetha, N., Karhade, J., Jatavallabhula, K.M., Yang, G., Scherer, S., Ramanan, D., Luiten, J.: Splatam: Splat, track & map 3d gaussians for dense rgb-d slam. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024) AHOY: Animatable Humans under Occlusion from YouTube Videos 19

    2024

  58. [58]

    ACM Transactions on Graphics42(4) (2023)

    Kerbl, B., Kopanas, G., Leimkühler, T., Drettakis, G.: 3d gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics42(4) (2023)

    2023

  59. [59]

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

    Khirodkar, R., Bagautdinov, T., Martinez, J., Zhaoen, S., James, A., Selednik, P., Anderson, S., Saito, S.: Sapiens: Foundation for human vision models. In: European Conference on Computer Vision (ECCV) (2024)

    2024

  60. [60]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    Kocabas, M., Chang, J.H.R., Gabriel, J., Tuzel, O., Ranjan, A.: HUGS: Human gaussian splats. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    2024

  61. [61]

    In: Advances in Neural Information Processing Systems (NeurIPS) (2023)

    Kolotouros, N., Alldieck, T., Zanfir, A., Bazavan, E., Fieraru, M., Sminchisescu, C.: DreamHuman: Animatable 3D avatars from text. In: Advances in Neural Information Processing Systems (NeurIPS) (2023)

    2023

  62. [62]

    arXiv preprint arXiv:2412.03603 (2024)

    Kong, W., et al.: HunyuanVideo: A systematic framework for large video generative models. arXiv preprint arXiv:2412.03603 (2024)

    Pith/arXiv arXiv 2024

  63. [63]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2021)

    Lassner, C., Zollhöfer, M.: Pulsar: Efficient sphere-based neural rendering. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2021)

    2021

  64. [64]

    In: Advances in Neural Information Processing Systems (NeurIPS) (2024)

    Lee, J., Won, C., Jung, H., Bae, I., Jeon, H.G.: Fully explicit dynamic gaussian splatting. In: Advances in Neural Information Processing Systems (NeurIPS) (2024)

    2024

  65. [65]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    Lei, J., Wang, Y., Pavlakos, G., Liu, L., Daniilidis, K.: GART: Gaussian articulated template models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    2024

  66. [66]

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

    Li, P., Zheng, W., Liu, Y., Yu, T., Li, Y., Qi, X., Chi, X., Xia, S., Cao, Y.P., Xue, W., Luo, W., Guo, Y.: Pshuman: Photorealistic single-image 3D human reconstruction using cross-scale multiview diffusion and explicit remeshing. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2025)

    2025

  67. [67]

    In: European Conference on Computer Vision (ECCV) (2022)

    Li, R., Tanke, J., Vo, M., Zollhofer, M., Gall, J., Kanazawa, A., Lassner, C.: Tava: Template-free animatable volumetric actors. In: European Conference on Computer Vision (ECCV) (2022)

    2022

  68. [68]

    ACM Transactions on Graphics (SIGGRAPH Asia) (2017)

    Li, T., Bolkart, T., Black, M.J., Li, H., Romero, J.: Learning a model of facial shape and expression from 4D scans. ACM Transactions on Graphics (SIGGRAPH Asia) (2017)

    2017

  69. [69]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    Li, Z., Chen, Z., Li, Z., Xu, Y.: Spacetime gaussian feature splatting for real- time dynamic view synthesis. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    2024

  70. [70]

    arXiv preprint arXiv:2311.16096 (2024)

    Li, Z., Sun, Y., Zheng, Z., Wang, L., Zhang, S., Liu, Y.: Animatable and relightable gaussians for high-fidelity human avatar modeling. arXiv preprint arXiv:2311.16096 (2024)

    Pith/arXiv arXiv 2024

  71. [71]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    Li, Z., Zheng, Z., Wang, L., Liu, Y.: Animatable gaussians: Learning pose- dependent gaussian maps for high-fidelity human avatar modeling. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)

    2024

  72. [72]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    Lin, G., Jiang, J., Yang, J., Zheng, Z., Liang, C.: OmniHuman-1: Rethinking the scaling-up of one-stage conditioned human animation models. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    2025

  73. [73]

    ACM Transactions on Graphics (2021)

    Liu, L., Habermann, M., Rudnev, V., Sarkar, K., Gu, J., Theobalt, C.: Neural actor: Neural free-view synthesis of human actors with pose control. ACM Transactions on Graphics (2021)

    2021

  74. [74]

    ACM Transactions on Graphics34(6) (2015) 20 A

    Loper, M., Mahmood, N., Romero, J., Pons-Moll, G., Black, M.J.: SMPL: A skinned multi-person linear model. ACM Transactions on Graphics34(6) (2015) 20 A. Mir et al

    2015

  75. [75]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    Lu, Y., Dong, J., Kwon, Y., Zhao, Q., Dai, B., De la Torre, F.: GAS: Gener- ative avatar synthesis from a single image. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    2025

  76. [76]

    In: International Conference on 3D Vision (3DV) (2024)

    Luiten, J., Kopanas, G., Leibe, B., Ramanan, D.: Dynamic 3d gaussians: Tracking by persistent dynamic view synthesis. In: International Conference on 3D Vision (3DV) (2024)

    2024

  77. [77]

    In: Pro- ceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    Luo, Y., Rong, Z., Wang, L., Zhang, L., Hu, T., Zhu, Y.: DreamActor-M1: Holistic, expressive and robust human image animation with hybrid guidance. In: Pro- ceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    2025

  78. [78]

    In: ACM SIGGRAPH Asia Conference Proceedings (2024)

    Ma, Y., Liu, H., Wang, H., Pan, H., He, Y., Yuan, J., Zeng, A., Cai, C., Shum, H.Y., Liu, W., Chen, Q.: Follow-your-emoji: Fine-controllable and expressive freestyle portrait animation. In: ACM SIGGRAPH Asia Conference Proceedings (2024)

    2024

  79. [79]

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

    Men, Y., Yao, Y., Cui, M., Bo, L.: MIMO: Controllable character video synthesis with spatial decomposed modeling. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2025)

    2025

  80. [80]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2019)

    Mescheder, L., Oechsle, M., Niemeyer, M., Nowozin, S., Geiger, A.: Occupancy networks: Learning 3d reconstruction in function space. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2019)

    2019

Showing first 80 references.