mlsanityData to 14 Oct 2026

Revisiting Avatar-As-Image: High-Fidelity Registration is All You Need

arXiv 2609.11722 ↗First posted 10 Sep 2026cs.CVcited by 1

Not on the map. Why?

Abstract (arXiv)

The representation of 3D clothed humans as standardized 2D UV texture and displacement maps over an underlying body model has long been studied. This compact representation is enticing as it enables pretrained image networks to process, generate, and edit 3D avatars, but is only useful if scans are accurately aligned and brought into correspondence via high-fidelity registration. This prerequisite has never been met, which we argue explains the limited quality of prior UV-based methods for clothed humans. Despite its significance, no public method produces high-fidelity SMPL(-X)+D registrations with UV texture from arbitrary clothed scans. We present AvaImg, a multi-stage optimization pipeline, to close this gap: it enforces body-inside-clothing constraint via signed winding numbers, made viable by a three-level efficiency cascade (~10x runtime reduced, ~95% storage saved), and recovers fine surface detail using coarse-to-fine displacement optimization. AvaImg outperforms all baselines in body fitting, shape estimation, and surface registration across six datasets, yielding textured registrations near-indistinguishable from scans (PSNR=34.48dB). For validation of AvaImg's Avatar-as-Image representation as imminently compatible with image foundation models, we auto-encode our UV maps via the frozen FLUX VAE. This achieves only 0.76mm added Chamfer error relative to scan and shows that the resulting maps lie within natural-image distributions, supporting the use of 2D generative priors for 3D avatar generation. Code, data, and Singularity containers will be at https://yuxuan-xue.com/avaimg.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 40 of its 54 references that are in the map. First posted on arXiv 10 Sep 2026.

Builds on

40 of 54 references.

References · 40

  1. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
  2. 2019Expressive Body Capture: 3D Hands, Face, and Body from a Single Imagecited by 1,612
  3. 2018OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fieldscited by 1,162
  4. 2020Accelerating 3D Deep Learning with PyTorch3Dcited by 784
  5. 2016Keep it SMPL: Automatic Estimation of 3D Human Pose and Shape from a Single Imagecited by 695
  6. 2021Function4D: Real-time Human Volumetric Capture from Very Sparse Consumer RGBD Sensorscited by 246
  7. 2019Multi-Garment Net: Learning to Dress 3D People from Imagescited by 245
  8. 2018Video Based Reconstruction of 3D People Modelscited by 238
  9. 2019DeepHuman: 3D Human Reconstruction from a Single Imagecited by 182
  10. 2019Tex2Shape: Detailed Full Human Body Geometry From a Single Imagecited by 166
  11. 2020Combining Implicit Function Learning and Parametric Models for 3D Human Reconstructioncited by 133
  12. 2021Real-time Deep Dynamic Characterscited by 133
28 more
  1. 2018Detailed Human Avatars from Monocular Videocited by 109
  2. 2020LoopReg: Self-supervised Learning of Implicit Surface Correspondences, Pose and Shape for 3D Human Mesh Registrationcited by 108
  3. 2023ASH: Animatable Gaussian Splats for Efficient and Photoreal Human Renderingcited by 104
  4. 2018DeepWrinkles: Accurate and Realistic Clothing Modelingcited by 90
  5. 2017Detailed, accurate, human shape estimation from clothed 3D scan sequencescited by 89
  6. 2023Paint-it: Text-to-Texture Synthesis via Deep Convolutional Texture Map Optimization and Physically-Based Renderingcited by 84
  7. 2019360-Degree Textures of People in Clothing from a Single Imagecited by 75
  8. 2023High-fidelity 3D Human Digitization from Single 2K Resolution Imagescited by 70
  9. 2020Learning to Transfer Texture from Clothing Images to 3D Humanscited by 58
  10. 2023Learning Locally Editable Virtual Humanscited by 54
  11. 20244D-DRESS: A 4D Dataset of Real-world Human Clothing with Semantic Annotationscited by 54
  12. 2024Human-3Diffusion: Realistic Avatar Creation via Explicit 3D Consistent Diffusion Modelscited by 44
  13. 2021Locally Aware Piecewise Transformation Fields for 3D Human Mesh Registrationcited by 41
  14. 2022Learned Vertex Descent: A New Direction for 3D Human Model Fittingcited by 37
  15. 2024An Object is Worth 64x64 Pixels: Generating 3D Object via Image Diffusioncited by 34
  16. 2023SMPLitex: A Generative Model and Dataset for 3D Human Texture Estimation from Single Imagecited by 30
  17. 2023TriHuman : A Real-time and Controllable Tri-plane Representation for Detailed Human Geometry and Appearance Synthesiscited by 28
  18. 2021Semi-supervised Synthesis of High-Resolution Editable Textures for 3D Humanscited by 21
  19. 2023NSF: Neural Surface Fields for Human Modeling from Monocular Depthcited by 21
  20. 2023NICP: Neural ICP for 3D Human Registration at Scalecited by 21
  21. 2024Gen-3Diffusion: Realistic Image-to-3D Generation via 2D & 3D Diffusion Synergycited by 16
  22. 2023Generalizing Neural Human Fitting to Unseen Poses With Articulated SE(3) Equivariancecited by 14
  23. 2025ETCH: Generalizing Body Fitting to Clothed Humans via Equivariant Tightnesscited by 9
  24. 2023SCULPT: Shape-Conditioned Unpaired Learning of Pose-dependent Clothed and Textured Human Meshescited by 7
  25. 2024Real-time Free-view Human Rendering from Sparse-view RGB Videos using Double Unprojected Texturescited by 7
  26. 2025Generative Human Geometry Distributioncited by 5
  27. 2026GeoRelight: Learning Joint Geometrical Relighting and Reconstruction with Flexible Multi-Modal Diffusion Transformerscited by 3
  28. 2026OmniFit: Multi-modal 3D Body Fitting via Scale-agnostic Dense Landmark Predictioncited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026COSMI: COmpositional Synthesis of Multi-object Interactionscited by 0