Pith. sign in

REVIEW 2 major objections 4 minor 45 references

Im2Haircut: Single-view Strand-based Hair Reconstruction for Human Avatars

T0 review · 2 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read A single photograph can be turned into strand-level 3D hair geometry in about ten minutes, with plausible occluded inner and back structure.

desk verdict The hybrid synthetic+real training scheme is a real step forward for single-view strand hair, but the synthetic 3D quantitative comparison is contaminated by PCA leakage on the test set, so the reported superiority is not yet proven. read the letter →

arxiv 2509.01469 v1 pith:SKLK4ME3 submitted 2025-09-01 cs.CV

Vanessa Sklyarova , Egor Zakharov , Malte Prinzler , Giorgio Becherini , Michael J. Black , Justus Thies This is my paper
classification cs.CV
keywords 3Dhairreconstructionstrand-basedsingle-viewhairstylepriorPCAmapGaussiansplattingsyntheticandrealtrainingdatahumanavatars
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

This paper claims that a single photograph is enough to reconstruct a full strand-based 3D hairstyle, including plausible geometry for the occluded inner and back regions. To get there, it learns a global hairstyle prior from a mixture of synthetic hairstyles with full 3D ground truth and real photographs with only 2D supervision, and then fine-tunes that prior against the input image with differentiable rendering. The authors report the lowest chamfer distances against synthetic ground truth and the best silhouette and orientation agreement among the compared single-view pipelines, with reconstructions ready in about ten minutes. If correct, this would make strand-level hair reconstruction practical for consumer avatar creation and animation, not just offline production.

What carries the argument

The load-bearing object is the PCA hair map: hair strands are compressed into 64 Fourier-domain PCA coefficients per scalp texel, with the first 10 coefficients forming a coarse branch and the remaining 54 a fine branch. A vision-transformer encoder-decoder predicts these maps, conditioned on orientation, segmentation, depth, and body-mask inputs; strand-aligned 3D Gaussians render the decoded strands differentiably. That renderer is what lets real photographs supervise the prior without 3D ground truth, through segmentation, direction, depth, and penetration losses, and it is also what drives the per-image fine-tuning at inference.

What would settle it

Fit the PCA strand basis on a held-out subset of the synthetic hairstyles and rerun the chamfer evaluation on hairdos excluded from that fit; if the reported errors do not survive, the benchmark is measuring basis memorization rather than reconstruction. Separately, compare against CT or multi-view strand ground truth for real heads to test the real-image claims.

Watch

Extended reading notes

Core claim

The central claim is that strand-based hair geometry can be recovered from one image by combining a global, low-dimensional hairstyle prior with local, rendering-based optimization. The prior is a transformer that predicts a 64-coefficient PCA hair map on the scalp UV space, split into 10 coarse coefficients and 54 fine ones; it is trained first on synthetic data with 3D losses, then jointly on real data using differentiable rendering. At test time, the predicted map becomes an initialization that is refined for 400 steps with losses on silhouette, direction, depth, and head penetration. The paper argues that this two-part design—hybrid prior plus inversion—captures detailed visible strand o

Load-bearing premise

The synthetic benchmark numbers assume the test hairstyles were not used to build the compact strand representation; the appendix says that representation was fit on all hairstyles from the same dataset the test set is drawn from.

Editorial extensions

If this is right

  • A single photograph is enough to produce a strand-based 3D hairstyle with plausible occluded inner and back geometry, in about 10 minutes on an A100.
  • Mixing synthetic 3D supervision with real-image rendering losses improves silhouette coverage and direction accuracy on real portraits compared with synthetic-only training.
  • Fine-tuning the prior with differentiable rendering recovers wavy structures and view alignment that pure regression misses.
  • The same prior supports multi-view reconstruction in about 45 minutes, an order of magnitude faster than the 10-hour direct optimization baseline.
  • The resulting strands are exportable to standard CG pipelines and can be simulated, making avatars animatable.

Reading between the lines

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

  • Not claimed by the paper: the PCA-basis overlap means the synthetic benchmark margin is likely optimistic; a held-out basis would test that.
  • Not claimed by the paper: the coarse-branch-only gradient rule could generalize as a stabilizer for hybrid training of other structural priors.
  • Not claimed by the paper: replacing the off-the-shelf direction estimator with one fine-tuned on real data could further improve backside consistency.
  • Not claimed by the paper: prior-space optimization may transfer to other avatar components where full 3D supervision exists only synthetically.
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. The paper presents Im2Haircut, a method for strand-based 3D hair reconstruction from a single photograph (and, with adaptation, from multiple views). The approach learns a hairstyle prior from a mixture of synthetic PERM data (with 3D supervision) and real Hairstep images (with self-supervised 3D Gaussian Splatting rendering losses). A transformer-based network predicts a 64-component PCA hair map in a coarse-to-fine manner, and at inference time the pretrained model is fine-tuned for 400 steps using rendering-based losses to align with the input image. The method is compared qualitatively and quantitatively against NeuralHDHair, Hairstep, PERM, HairNet, and Hairmony, and the paper claims superior performance in capturing hair orientation, silhouette, and backside consistency.

Significance. If the quantitative claims are validated, the paper would be a useful contribution to single-view hair reconstruction: it combines a learned global hair prior with local optimization, uses both synthetic and real data to reduce the domain gap, and demonstrates applications in multi-view reconstruction and simulation. The pipeline is relatively fast (about 10 minutes on an A100) and the authors state that code will be released. The main weakness is that the headline synthetic comparison is currently compromised by a PCA basis leakage in the evaluation setup, as detailed below.

major comments (2)
  1. [S2.1, Sec. 5.1, Eq. (1), Table 1] The quantitative synthetic evaluation is undermined by PCA basis leakage. Supplementary S2.1 states that the Incremental PCA basis is computed on 'all hairstyles from the PERM dataset', and Sec. 5.1 evaluates on ten PERM hairstyles. Since Eq. (1) decodes every reconstruction as iDFT(\bar{S} + \gamma^T X), all outputs are constrained to the 64-dimensional subspace spanned by the PERM strands. If the test hairstyles are included in the PCA fit, the representation space already contains the test geometry, which selectively lowers the reported chamfer distances for the proposed method relative to baselines whose output spaces are not so tuned. The reported difference to NeuralHDHair is very small (chamferpts 0.000246 vs 0.000251), so the leakage could change the ranking. Please re-run the PCA basis on the training split only (or use leave-one-out over the ten test hairstyles), report per-hai
  2. [Table 1, Sec. 5.1] The statistical support for the 'superior performance' claim is weak. The synthetic comparison uses only 10 hairstyles and the real-image comparison uses 50 images, with no error bars, no per-sample distributions, and no significance tests. Given the small effect size in the synthetic chamfer comparison, the claim of superiority would be much more convincing if the authors reported per-hairstyle chamfer values, standard deviations, and a paired test (e.g., Wilcoxon signed-rank or bootstrap confidence intervals). Without this, the quantitative ranking is not established.
minor comments (4)
  1. [Sec. 5.1, References] The method list refers to 'Hairmony [43]' but the correct reference number is [24]. The reference list contains Hairmony as [24]; please fix the citation.
  2. [Sec. 5.2, Table 2] The metric 'mask↓' is used in Tables 2 and 3 but is not defined in the main text. Please define the mask error and specify whether it is an L2 distance or IoU-based score.
  3. [Sec. 5.1, Eq. (12)] The Lundir metric is used in Table 1 but only defined later in Eq. (12). Move the definition earlier or provide a pointer at first use.
  4. [S3, Table 1] The synthetic evaluation says metrics are computed on 10,000 sampled strands; please clarify the sampling procedure and whether the same strand count is used for all methods. Also, note that the hardware used for the 10-minute timing (A100) differs from the RTX3090 used for Hairstep; a like-for-like timing would improve comparability.

Circularity Check

1 steps flagged · score 6.0 of 10

Synthetic 3D comparison is partially circular: the PCA basis fit on all PERM hairstyles includes the test hairstyles, so Table 1's chamfer advantage partly reflects test-set leakage.

  1. fitted input called prediction [Section S2.1 (strand PCA basis) and Section 5.1 / Table 1 (synthetic quantitative comparison)]
    "For basis calculation, we launch the Incremental PCA method on all hairstyles from the PERM [11] dataset. ... We use ten hairstyles and render them from a hemisphere with high-resolution textures to reduce the domain gap between synthetic renders and real images. We launch Hairstep [42], NeuralHDHair [36], and our method on the obtained data, align results in the same space, and calculate the chamfer distance on points and directions between reconstructed and ground-truth 3D hairstyles; see Table 1 for results."

    Equation (1) decodes every prediction as iDFT(mean + γ^T X), so all outputs are confined to the 64-dimensional subspace spanned by the PERM strands used to fit X. Since S2.1 fits X on all PERM hairstyles and Section 5.1 evaluates on ten PERM hairstyles from the same dataset, the reconstruction space already contains the test geometry. The chamfer comparison then measures reconstruction inside a test-informed subspace against baselines (NeuralHDHair, Hairstep) whose output spaces were not fit to those same test strands. This selectively lowers chamferpts and chamferangle for Im2Haircut; the 0.000246 vs 0.000251 margin over NeuralHDHair is too small and too confounded to support the quantitative superiority claim. The real-image metrics and qualitative back-view comparisons do not share this

full rationale

The paper's core method is an empirical reconstruction pipeline, not a formal derivation, so most of the chain (transformer coarse/fine branches, Gaussian-splatting losses, single-view inversion) is self-contained and does not reduce to its inputs. The one concrete circular step is the synthetic quantitative evaluation: the PCA basis X in Eq. (1) is fit by Incremental PCA on 'all hairstyles from the PERM dataset' (S2.1), and the 3D test set in Section 5.1 is ten PERM hairstyles. Because Eq. (1) writes every prediction as iDFT(mean + γ^T X), all predictions lie in the subspace spanned by the test strands. This does not by itself make the network output equal to ground truth, but it removes the representation bottleneck for the tested geometry while the baselines are evaluated in their own, not test-tuned, output spaces, biasing the chamfer comparison in Table 1. The qualitative comparisons, real-image metrics (IoU/Lundir on 50 Hairstep images), and ablations are independent of this particular leakage, so the central claim retains independent content; the score is therefore partial, not total. I did not find load-bearing self-citation circularity: references to prior work [11, 38, 33] are used as components or representations, not as an unverified uniqueness theorem forcing the result.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the PERM synthetic distribution being representative, on a PCA basis that is fit to the full PERM dataset, and on trust in off-the-shelf estimators for depth, segmentation, direction, and camera. The hand-set loss weights and mixing rate tune the training objective; no new physical entities are introduced.

free parameters (7)
  • PCA basis X = 64 components fit to all 21,054 PERM hairstyles
    Defines the strand representation (Eq. 1); computed on the full dataset including evaluation subjects, per Supp. S2.1.
  • PCA coefficient count gamma = 64
    Chosen following PERM/GroomGen [11,44]; sets the capacity of the hair map.
  • Loss weights (lambda) = coarse: lambda_PCA=0.1, lambda_dir=0.1, lambda_curv=1, lambda_mask=1e-4; fine: lambda_PCA=10, lambda_dir=0.1, lambda_cur
    Hand-set hyperparameters that trade off geometry, orientation, segmentation, and penetration terms; ablations show directional and silhouette losses are critical (Tables 4,5).
  • Mixing rate r = 0.5
    Weight of real-data loss in hybrid training (Eq. 11); chosen by hand, impacts realism vs synthetic accuracy.
  • Visibility weight wij = 3 for visible points, 1 otherwise
    Weighted MAE focuses fine stage on visible strands following HairNet [43].
  • Gaussian width values = epsilon=0.005 during training, 0.00035 during inversion
    Affects differentiable rendering tightness and optimization behavior, set by hand.
  • Number of strands = ~10,000 during optimization, 30,000 for visualization
    Chosen for quality/cost tradeoff; comparisons use different strand counts across baselines.
assumptions (5)
  • domain assumption PERM synthetic hairstyles are a valid proxy for real hairstyle geometry and statistics
    The PCA basis and synthetic training loss are built entirely from PERM; the paper does not validate this against external hair scans (Sec. 4.1, S2.1).
  • domain assumption The FLAME head UV parametrization and scalp region describe the hair root space for real subjects
    Hair maps are predicted on FLAME UV; real images are aligned via facial keypoints to the same bust model (Supp. S2.3).
  • domain assumption Strand-aligned 3D Gaussian splatting is a faithful differentiable renderer for hair appearance
    Used for real-data self-supervision and inversion (Eq. 13, Sec. 4.1); the validity is inherited from GaussianHaircut [38].
  • domain assumption Off-the-shelf estimators (Depth Pro, SAM, Hairstep direction maps, Deep3DFace) provide errors small enough not to break reconstruction
    The paper lists this as a limitation in Sec. 5.4 but still loads these inputs into the training and inference pipeline.
  • standard math Incremental PCA and inverse DFT are standard operations; the chosen 64 components capture strand geometry
    Eq. 1 and Supp. S2.1 assert the representation; no error bound between original and PCA-reconstructed strands is reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Im2Haircut: Single-view Strand-based Hair Reconstruction for Human Avatars." pith.science (2026). https://pith.science/paper/SKLK4ME3

@misc{pith2026250901469,
  author       = {Pith},
  title        = {Pith review of: Im2Haircut: Single-view Strand-based Hair Reconstruction for Human Avatars},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SKLK4ME3}},
  note         = {Machine review of arXiv:2509.01469}
}
read the original abstract

We present a novel approach for 3D hair reconstruction from single photographs based on a global hair prior combined with local optimization. Capturing strand-based hair geometry from single photographs is challenging due to the variety and geometric complexity of hairstyles and the lack of ground truth training data. Classical reconstruction methods like multi-view stereo only reconstruct the visible hair strands, missing the inner structure of hairstyles and hampering realistic hair simulation. To address this, existing methods leverage hairstyle priors trained on synthetic data. Such data, however, is limited in both quantity and quality since it requires manual work from skilled artists to model the 3D hairstyles and create near-photorealistic renderings. To address this, we propose a novel approach that uses both, real and synthetic data to learn an effective hairstyle prior. Specifically, we train a transformer-based prior model on synthetic data to obtain knowledge of the internal hairstyle geometry and introduce real data in the learning process to model the outer structure. This training scheme is able to model the visible hair strands depicted in an input image, while preserving the general 3D structure of hairstyles. We exploit this prior to create a Gaussian-splatting-based reconstruction method that creates hairstyles from one or more images. Qualitative and quantitative comparisons with existing reconstruction pipelines demonstrate the effectiveness and superior performance of our method for capturing detailed hair orientation, overall silhouette, and backside consistency. For additional results and code, please refer to https://im2haircut.is.tue.mpg.de.

Figures

Figures reproduced from arXiv: 2509.01469 by the authors.

Figure 1
Figure 1. Given a single image, our method, Im2Haircut, generates high-quality, strand-based 3D hair geometry. Im2Haircut consists of a [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Our inference pipeline consists of two stages: coarse and fine. We first train to predict the first 10 components of the PCA hair [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Comparison results with baseline. NeuralHDHair and Hairstep produce smooth geometry, while Hairstep also has artifacts from the back view of short hairstyles (see row 4). PERM reconstructs simple hairstyles well (see row 1), but struggles with more complicated geometry. Our method recovers more detailed hairstyles than others. Digital zoom-in is recommended. Quantitative comparison. First, we perform quantitative co… view at source ↗
Figures from the paper (23 more)
Figure 6
Figure 6. Figure 6: Comparison of our method with HairNet [43]. 5.2. Ablation studies We evaluate our losses, the influence of the depth map signal on reconstruction quality, as well as the training setup, using the following metrics: a chamfer distance with normals between predicted and …
Figure 5
Figure 5. Figure 5: Out-of-distribution comparison of our method (second column) with Hairstep [42] (last column). Qualitative comparison [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 7
Figure 7. Figure 7: Hair reconstruction stage. The mixing strategy and prior space are important for single-view inversion. chamfer pts ↓ chamfer angle ↓ angle error ↓ mask ↓ Lundir ↓ hybrid training 0.00030 0.143 18.03 0.405 0.695 w/o coarse-to-fine 0.00032 0.135 17.87 0.578 0.732 w/o Mi…
Figure 11
Figure 11. Figure 11: Combination of Gabor map with Direction map from [PITH_FULL_IMAGE:figures/full_fig_p011_11.png]
Figure 12
Figure 12. Figure 12: Synthetic dataset is generated by rendering ground￾truth hairstyles in Blender. ralHDHair [36] and PERM [11] were provided by the re￾spective authors. For Hairnet [43], we use images extracted from the Hairstep [42] paper, and do not compare on more samples, as the pe…
Figure 13
Figure 13. Figure 13: Comparison of our model with retrieval-based method Hairmony [24] . Image Ground-truth w Mixing w/o Mixing [PITH_FULL_IMAGE:figures/full_fig_p013_13.png]
Figure 14
Figure 14. Figure 14: Mixing strategy. The performance of the Hairstyle prior model with and without mixing strategy. The model trained on real images can better regress the hair silhouette and orienta￾tions. The color in the image corresponds to the direction in the orientation maps. retr…
Figure 15
Figure 15. Figure 15: Hair reconstruction stage. We demonstrate the importance of optimizing in Hairstyle prior space of the model for single-view inversion (compare “Ours” to “w/o prior3D” and “w/o priorpca”). Also, we show a scenario when only the decoder is optimized while the encoder i…
Figure 16
Figure 16. Figure 16: Comparison with Hairstep under the same computa￾tional cost. chamfer pts ↓ chamfer angle ↓ angle error ↓ mask ↓ Lundir ↓ hybrid training 0.00030 0.143 18.03 0.405 0.695 Priorw/o dir 0.00028 0.140 17.84 0.418 0.723 Priorw/o penetr 0.00030 0.139 17.60 0.412 0.689 Priorw…
Figure 17
Figure 17. Figure 17: Extended ablation on losses. We show the importance of mixing strategy by training a prior model using rendering loss computed only on synthetic data (see Priorsyn only). Also, we provide an ablation on losses during hybrid model training with post optimiza￾tion using…
Figure 18
Figure 18. Figure 18: Back view comparison. Comparison of back view geometry of reconstructions obtained by our method, Hairstep [42] and NeuralHDHair [36]. 16 [PITH_FULL_IMAGE:figures/full_fig_p016_18.png]
Figure 19
Figure 19. Figure 19: Extended ablation on importance of training on synthetic and real data. Results of “Ours” correspond to columns 2-4, while “w/o Mixing” to 5-7. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_19.png]
Figure 20
Figure 20. Figure 20: Extended ablation on importance of training on synthetic and real data. Results of “Ours” correspond to columns 2-4, while “w/o Mixing” to 5-7. 18 [PITH_FULL_IMAGE:figures/full_fig_p018_20.png]
Figure 21
Figure 21. Figure 21: Extended qualitative comparison using real-world multi-view scenes [28] with Gaussian Haircut (GH) [38]. We compared in a scenario with 1, 3, 8, and 32 views available. Note, GH fails in scenarios with 1 and 3 views. Digital zoom-in is recommended. 19 [PITH_FULL_IMAG…
Figure 22
Figure 22. Figure 22: Extended qualitative comparison of our method (columns 2–4) with Hairstep [42] (last three columns) on wavy samples. Our method can reconstruct curlier structures with more realistic back geometry. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_22.png]
Figure 23
Figure 23. Figure 23: Extended qualitative comparison of our method (columns 2–4) with Hairstep [42] (last three columns) on out-of￾distribution samples. 21 [PITH_FULL_IMAGE:figures/full_fig_p021_23.png]
Figure 24
Figure 24. Figure 24: More results of our method on out-of-distribution data. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_24.png]
Figure 25
Figure 25. Figure 25: More results of our method on out-of-distribution data. 23 [PITH_FULL_IMAGE:figures/full_fig_p023_25.png]
Figure 26
Figure 26. Figure 26: Additional results of our model. 24 [PITH_FULL_IMAGE:figures/full_fig_p024_26.png]
Figure 27
Figure 27. Figure 27: Additional results of our model. 25 [PITH_FULL_IMAGE:figures/full_fig_p025_27.png]
Figure 28
Figure 28. Figure 28: Additional results of our model. 26 [PITH_FULL_IMAGE:figures/full_fig_p026_28.png]
Figure 29
Figure 29. Figure 29: Extended qualitative comparison with Hairstep [42] and NeuralHDHair [36]. Note that in NeuralHDHair the number of rendered strands is twice as compared to our method and Hairstep. 27 [PITH_FULL_IMAGE:figures/full_fig_p027_29.png]
Figure 30
Figure 30. Figure 30: Extended qualitative comparison with Hairstep [42] and NeuralHDHair [36]. 28 [PITH_FULL_IMAGE:figures/full_fig_p028_30.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

45 extracted references · 40 canonical work pages

  1. [1]

    Discrete elastic rods

    Mikl ´os Bergou, Max Wardetzky, Stephen Robinson, Basile Audoly, and Eitan Grinspun. Discrete elastic rods. ACM SIGGRAPH 2008 papers, 2008. 2

    work page 2008

  2. [2]

    Richter, and Vladlen Koltun

    Aleksei Bochkovskii, Ama ¨el Delaunoy, Hugo Germain, Marcel Santos, Yichao Zhou, Stephan R. Richter, and Vladlen Koltun. Depth pro: Sharp monocular metric depth in less than a second. ArXiv, abs/2410.02073, 2024. 5, 11

    arXiv 2024

  3. [3]

    Single-view hair modeling for portrait manipulation

    Menglei Chai, Lvdi Wang, Yanlin Weng, Yizhou Yu, Baining Guo, and Kun Zhou. Single-view hair modeling for portrait manipulation. ACM Transactions on Graphics (TOG), 31:1 – 8, 2012. 2

    work page 2012

  4. [4]

    Autohair.ACM Transactions on Graphics (TOG), 35:1 – 12, 2016

    Menglei Chai, Tianjia Shao, Hongzhi Wu, Yanlin Weng, and Kun Zhou. Autohair.ACM Transactions on Graphics (TOG), 35:1 – 12, 2016. 2

    work page 2016

  5. [5]

    A practical and controllable hair and fur model for production path tracing

    Matt Jen-Yuan Chiang, Benedikt Bitterli, Chuck Tappan, and Brent Burley. A practical and controllable hair and fur model for production path tracing. Computer Graphics Forum, 35,

    work page

  6. [6]

    Blender - a 3D modelling and rendering package

    Blender Online Community. Blender - a 3D modelling and rendering package. Blender Foundation, Stichting Blender Foundation, Amsterdam, 2018. 5, 12

    work page 2018

  7. [7]

    Interactive hair simulation on the gpu using admm

    Gilles Daviet. Interactive hair simulation on the gpu using admm. ACM SIGGRAPH 2023 Conference Proceedings ,

    work page 2023

  8. [8]

    Accurate 3d face reconstruction with weakly-supervised learning: From single image to image set

    Yu Deng, Jiaolong Yang, Sicheng Xu, Dong Chen, Yunde Jia, and Xin Tong. Accurate 3d face reconstruction with weakly-supervised learning: From single image to image set. In IEEE Computer Vision and Pattern Recognition Work- shops, 2019. 5

    work page 2019

Show all 45 references
  1. [9]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition ...

    2010 arXiv

  2. [10]

    Unreal engine

    Epic Games. Unreal engine. 12

  3. [11]

    Perm: A parametric representation for multi-style 3D hair modeling

    Chengan He, Xin Sun, Zhixin Shu, Fujun Luan, S ¨oren Pirk, Jorge Alejandro Amador Herrera, Dominik L Michels, Tuan- feng Y Wang, Meng Zhang, Holly Rushmeier, and Yi Zhou. Perm: A parametric representation for multi-style 3D hair modeling. In International Conference on Learnin...

    2025

  4. [12]

    Single- view hair modeling using a hairstyle database

    Liwen Hu, Chongyang Ma, Linjie Luo, and Hao Li. Single- view hair modeling using a hairstyle database. ACM Trans- actions on Graphics (Proceedings of SIGGRAPH 2015) , 34 (4):125:1–125:9, 2015. 2, 12

    2015

  5. [13]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkuehler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics (TOG), 42:1 – 14, 2023. 2

    2023

  6. [14]

    3D gaussian splatting for real-time radiance field rendering

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

    2023

  7. [15]

    Berg, Wan-Yen Lo, Piotr Doll ´ar, and Ross B

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chlo´e Rolland, Laura Gustafson, Tete Xiao, Spencer White- head, Alexander C. Berg, Wan-Yen Lo, Piotr Doll ´ar, and Ross B. Girshick. Segment anything. 2023 IEEE/CVF In- ternational Conference on Computer Vision (ICCV) ...

    2023

  8. [16]

    DeepMVSHair: Deep hair modeling from sparse views

    Zhiyi Kuang, Yiyang Chen, Hongbo Fu, Kun Zhou, and Youyi Zheng. DeepMVSHair: Deep hair modeling from sparse views. In SIGGRAPH Asia 2022 Conference Papers, New York, NY , USA, 2022. Association for Computing Ma- chinery. 2

    2022

  9. [17]

    Tianye Li, Timo Bolkart, Michael. J. Black, Hao Li, and Javier Romero. Learning a model of facial shape and ex- pression from 4D scans. ACM Transactions on Graphics, (Proc. SIGGRAPH Asia), 36(6):194:1–194:17, 2017. 4

    2017

  10. [18]

    Mat: Mask-aware transformer for large hole image in- painting

    Wenbo Li, Zhe Lin, Kun Zhou, Lu Qi, Yi Wang, and Jiaya Jia. Mat: Mask-aware transformer for large hole image in- painting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022. 4

    2022

  11. [19]

    Tangled: Generating 3d hair strands from images with arbitrary styles and viewpoints

    Pengyu Long, Zijun Zhao, Min Ouyang, Qingcheng Zhao, Qixuan Zhang, Wei Yang, Lan Xu, and Jingyi Yu. Tangled: Generating 3d hair strands from images with arbitrary styles and viewpoints. arXiv preprint arXiv:2502.06392, 2025. 2

    2025 arXiv

  12. [20]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019. OpenReview.net, 2019. 5, 11

    2019

  13. [21]

    GaussianHair: Hair modeling and rendering with light-aware gaussians

    Haimin Luo, Min Ouyang, Zijun Zhao, Suyi Jiang, Longwen Zhang, Qixuan Zhang, Wei Yang, Lan Xu, and Jingyi Yu. GaussianHair: Hair modeling and rendering with light-aware gaussians. arXiv preprint arXiv:2402.10483, 2024. 2, 3

    2024 arXiv

  14. [22]

    Multi-view hair capture using orientation fields

    Linjie Luo, Hao Li, Sylvain Paris, Thibaut Weise, Mark Pauly, and Szymon Rusinkiewicz. Multi-view hair capture using orientation fields. 2012 IEEE Conference on Computer Vision and Pattern Recognition, pages 1490–1497, 2012. 2

    2012

  15. [23]

    Light scattering from hu- man hair fibers

    Steve Marschner, Henrik Wann Jensen, Mike Cammarano, Steven Worley, and Pat Hanrahan. Light scattering from hu- man hair fibers. ACM SIGGRAPH 2003 Papers, 2003. 2

    2003

  16. [24]

    Hairmony: Fairness-aware hairstyle classification

    Givi Meishvili, James Clemoes, Charlie Hewitt, Zafiirah Ho- senie, Xian Xiao, Martin de La Gorce, Tibor Takacs, Tadas 9 Baltrusaitis, Antonio Criminisi, Chyna McRae, Nina Jablon- ski, and Marta Wilczkowiak. Hairmony: Fairness-aware hairstyle classification. In SIGGRAPH Asia 20...

    2024

  17. [25]

    Kim, and Yaser Sheikh

    Giljoo Nam, Chenglei Wu, Min H. Kim, and Yaser Sheikh. Strand-accurate multi-view hair capture. In 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 155–164, 2019. 2

    2019

  18. [26]

    Brice ˜no, and Franc ¸ois X

    Sylvain Paris, H ´ector M. Brice ˜no, and Franc ¸ois X. Sillion. Capture of hair geometry from multiple images.ACM Trans- actions on Graphics (TOG), 23:712 – 719, 2004. 2

    2004

  19. [27]

    Kozhushnyan, Wo- jciech Jarosz, Wojciech Matusik, Matthias Zwicker, and Fr´edo Durand

    Sylvain Paris, William Chang, Oleg I. Kozhushnyan, Wo- jciech Jarosz, Wojciech Matusik, Matthias Zwicker, and Fr´edo Durand. Hair photobooth: geometric and photometric acquisition of real hairstyles.ACM SIGGRAPH 2008 papers,

    2008

  20. [28]

    H3d-net: Few-shot high-fidelity 3d head reconstruc- tion

    Eduard Ramon, Gil Triginer, Janna Escur, Albert Pumarola, Jaime Garcia, Xavier Giro-i Nieto, and Francesc Moreno- Noguer. H3d-net: Few-shot high-fidelity 3d head reconstruc- tion. In Proceedings of the IEEE/CVF International Confer- ence on Computer Vision , pages 5620–5629, 2...

    2021

  21. [29]

    Neural strands: Learning hair geometry and appearance from multi-view im- ages

    Radu Alexandru Rosu, Shunsuke Saito, Ziyan Wang, Chen- glei Wu, Sven Behnke, and Giljoo Nam. Neural strands: Learning hair geometry and appearance from multi-view im- ages. In Computer Vision – ECCV 2018: 15th European Conference, 2022. 2, 3

    2018

  22. [30]

    Radu Alexandru Rosu, Keyu Wu, Yao Feng, Youyi Zheng, and Michael J. Black. DiffLocks: Generating 3D hair from a single image using diffusion models. In Proceedings IEEE/CVF Conf. on Computer Vision and Pattern Recogni- tion (CVPR), 2025. 2, 3

    2025

  23. [31]

    Ct2hair: High-fidelity 3d hair modeling using com- puted tomography

    Yuefan Shen, Shunsuke Saito, Ziyan Wang, Olivier Maury, Chenglei Wu, Jessica Hodgins, Youyi Zheng, and Giljoo Nam. Ct2hair: High-fidelity 3d hair modeling using com- puted tomography. ACM Trans. Graph., 42(4), 2023. 2

    2023

  24. [32]

    Neural haircut: Prior-guided strand-based hair reconstruction

    Vanessa Sklyarova, Jenya Chelishev, Andreea Dogaru, Igor Medvedev, Victor Lempitsky, and Egor Zakharov. Neural haircut: Prior-guided strand-based hair reconstruction. In Proceedings of IEEE International Conference on Computer Vision (ICCV), 2023. 2, 3, 4

    2023

  25. [33]

    Black, and Justus Thies

    Vanessa Sklyarova, Egor Zakharov, Otmar Hilliges, Michael J. Black, and Justus Thies. Text-conditioned gen- erative model of 3d strand-based human hairstyles. 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 4703–4712, 2023. 5, 11

    2024

  26. [34]

    Dr.hair: Reconstructing scalp- connected hair strands without pre-training via differentiable rendering of line segments

    Yusuke Takimoto, Hikari Takehara, Hiroyuki Sato, Zi- hao Zhu, and Bo Zheng. Dr.hair: Reconstructing scalp- connected hair strands without pre-training via differentiable rendering of line segments. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogn...

    2024

  27. [35]

    OpenGL programming guide: the official guide to learning OpenGL, version 1.2

    Mason Woo, Jackie Neider, Tom Davis, and Dave Shreiner. OpenGL programming guide: the official guide to learning OpenGL, version 1.2. Addison-Wesley Longman Publishing Co., Inc., 1999. 6, 14

    1999

  28. [36]

    NeuralHDHair: Automatic high-fidelity hair modeling from a single image using implicit neural rep- resentations

    Keyu Wu, Yifan Ye, Lingchen Yang, Hongbo Fu, Kun Zhou, and Youyi Zhengl. NeuralHDHair: Automatic high-fidelity hair modeling from a single image using implicit neural rep- resentations. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , page...

    2022

  29. [37]

    Monohair: High-fidelity hair modeling from a monocular video

    Keyu Wu, Lingchen Yang, Zhiyi Kuang, Yao Feng, Xutao Han, Yuefan Shen, Hongbo Fu, Kun Zhou, and Youyi Zheng. Monohair: High-fidelity hair modeling from a monocular video. In 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024. 2

    2023

  30. [38]

    Human hair recon- struction with strand-aligned 3d gaussians

    Egor Zakharov, Vanessa Sklyarova, Michael Black, Giljoo Nam, Justus Thies, and Otmar Hilliges. Human hair recon- struction with strand-aligned 3d gaussians. In Computer Vi- sion – ECCV 2024: 18th European Conference, Milan, Italy, September 29–October 4, 2024, Proceedings, Par...

    2024

  31. [39]

    Hair-gans: Recovering 3d hair structure from a single image

    Meng Zhang and Youyi Zheng. Hair-gans: Recovering 3d hair structure from a single image. ArXiv, abs/1811.06229,

    arXiv

  32. [40]

    A data-driven approach to four-view image-based hair modeling

    Meng Zhang, Menglei Chai, Hongzhi Wu, Hao Yang, and Kun Zhou. A data-driven approach to four-view image-based hair modeling. ACM Transactions on Graphics (TOG), 36:1 – 11, 2017. 2

    2017

  33. [41]

    Modeling hair from an rgb-d camera

    Meng Zhang, Pan Wu, Hongzhi Wu, Yanlin Weng, Youyi Zheng, and Kun Zhou. Modeling hair from an rgb-d camera. ACM Transactions on Graphics (TOG), 37:1 – 10, 2018. 2

    2018

  34. [42]

    HairStep: Transfer synthetic to real using strand and depth maps for single-view 3d hair modeling

    Yujian Zheng, Zirong Jin, Moran Li, Haibin Huang, Chongyang Ma, Shuguang Cui, and Xiaoguang Han. HairStep: Transfer synthetic to real using strand and depth maps for single-view 3d hair modeling. In 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pa...

    2023

  35. [43]

    HairNet: Single-view hair recon- struction using convolutional neural networks

    Yi Zhou, Liwen Hu, Jun Xing, Weikai Chen, Han-Wei Kung, Xin Tong, and Hao Li. HairNet: Single-view hair recon- struction using convolutional neural networks. In Computer Vision – ECCV 2018: 15th European Conference, Munich, Germany, September 8-14, 2018, Proceedings, Part XI, ...

    2018

  36. [44]

    Groomgen: A high-quality gen- erative hair model using hierarchical latent representations

    Yuxiao Zhou, Menglei Chai, Alessandro Pepe, Markus Gross, and Thabo Beeler. Groomgen: A high-quality gen- erative hair model using hierarchical latent representations. ACM Trans. Graph., 42(6), 2023. 3

    2023

  37. [45]

    Ours same cost

    Yuxiao Zhou, Menglei Chai, Daoye Wang, Sebastian Win- berg, Erroll Wood, Kripasindhu Sarkar, Markus Gross, and Thabo Beeler. GroomCap: High-fidelity prior-free hair cap- ture. ACM Trans. Graph., 43(6), 2024. 2, 3 10 S1. Background To use 3D Gaussian Splatting for soft-rasteriz...

    2024

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.