Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

TANGLED: Generating 3D Hair Strands from Images with Arbitrary Styles and Viewpoints

T0 review · 3 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read The paper claims that conditioning a latent diffusion model on multi-view lineart lets one to eight images of any style generate 3D strand-based hair without per-domain retraining.

desk verdict Plausible method and a useful dataset, but the generalization claim is evaluated only on the authors' own rendered and stylized data, so the evidence is thinner than the title suggests. read the letter →

arxiv 2502.06392 v1 pith:CIWDROMV submitted 2025-02-10 cs.CV cs.GR

Pengyu Long , Zijun Zhao , Min Ouyang , Qingcheng Zhao , Qixuan Zhang , Wei Yang , Lan Xu , Jingyi Yu This is my paper · ORCID
classification cs.CVcs.GR
keywords 3Dhairgenerationlineartconditioninglatentdiffusionmodelmulti-viewreconstructionstrand-basedrepresentationbraidinpaintingMultidatasetculturallydiversehairstyles
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

TANGLED sets out to make 3D hair generation as flexible as image input: one to eight photos, drawings, or sketches in any style and viewpoint should yield a complete strand-based hairstyle ready for animation. The paper argues that the key is to condition a latent diffusion model not on the raw images but on their extracted lineart, because lineart preserves structural cues such as strand density and parting lines while discarding lighting, texture, and background noise. It supports this by introducing MultiHair, a dataset of 457 hairstyles with 74 annotations and rendered multi-view images, and by adding a parametric braid-inpainting step for structurally complex styles. If the central claim is right, diverse and culturally specific hairstyles such as braids, locs, and curly or afro textures can be produced without manual modeling or controlled capture rigs, and user studies reported a strong preference over text-guided and optimization baselines.

What carries the argument

The central object is the multi-view lineart condition: a sparse contour representation extracted by a lineart detector from hair-region images after mask segmentation. DINOv2 turns each lineart view into features, and random masking of one to eight views during training teaches the latent diffusion U-Net to accept arbitrary numbers of viewpoints. The diffusion operates on a 32x32 UV latent map of hair strands, using a strand-based polyline representation encoded by a variational autoencoder. For braids, a parametric module builds helical geometry around a guide curve, maps it to the latent space, and uses attention-based inpainting to replace the masked braid region during denoising.

What would settle it

Run the released TANGLED model on a held-out set of photographs and stylized images of braided, loc'd, and curly hairstyles that were never part of MultiHair, and compare against HairStep and HAAR with their official checkpoints on the same inputs; if the lineart-conditioned model does not preserve strand density and parting lines or does not beat the baselines outside its own test split, the central robustness claim fails.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that multi-view lineart is a sufficient and robust conditioning signal for generating 3D hair strands across arbitrary styles and viewpoints. A latent diffusion model is trained in a UV hair-latent space, with DINOv2 features of lineart images from one to eight views injected through cross-attention. The lineart representation is what lets the model transfer between photographs, anime, oil paintings, and sketches: the same hairstyle rendered in different visual domains yields similar lineart, so the model captures geometry and topology rather than appearance. A post-processing module further encodes braid geometry parametrically and inpaints it into the generated latent during denoising, preserving coherent crossings and torsion. The paper reports that this configuration beats text-guided and single-view optimization baselines on geometric and semantic metrics and is preferred by users, especially when multiple views are available.

Load-bearing premise

The load-bearing premise is that comparing pre-trained baselines on a 5% split of the authors' own MultiHair dataset is a fair and representative test of arbitrary styles and viewpoints, rather than a test that disadvantages methods never trained on that data.

Editorial extensions

If this is right

  • The same trained model accepts photos, anime, oil paintings, sketches, and partial renders because all input is reduced to lineart before conditioning.
  • Conditioning on four views rather than one removes occlusion-driven errors, such as failing to reconstruct the hidden side of asymmetric or shoulder-length hair.
  • Generated strands are polylines with scalp UV layout, so they can be exported into standard animation and rendering tools for editing and virtual production.
  • The MultiHair dataset's emphasis on braids, locs, and underrepresented textures is what allows the model to produce styles that earlier datasets under-represent.
  • Parametric braid inpainting prevents the topological errors, such as broken crossings and distorted torsion, that pure diffusion produces on complex braided styles.

Reading between the lines

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

  • Because lineart discards color and lighting, the paper's method assigns hair color manually; extending it to identity-preserving avatar generation would require a separate color-conditioning channel.
  • The same lineart-conditioned diffusion framework may transfer to other strand-like structures such as fur, eyebrows, or cloth fibers, where topology matters more than photorealistic appearance.
  • The training-time blending of features from one to eight views likely acts as a view-dropout regularizer, teaching cross-view consistency; measuring cross-view reconstruction error on external hairstyles would test this indirect benefit.
  • The reported superiority of lineart conditioning over raw-image conditioning suggests that the representation itself, rather than dataset size alone, drives the generalization gain across styles.
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

3 major / 5 minor

Summary. The paper presents TANGLED, a three-stage pipeline for generating 3D hair strand models from one or more input images of arbitrary artistic style and viewpoint. The authors introduce MultiHair, a dataset of 457 3D hairstyles with 74 attributes, 72 rendered views per hairstyle, and synthetic stylized images generated via ControlNet. The core model is a latent diffusion framework that conditions on DINOv2 features extracted from multi-view lineart images, with random view selection during training to support flexible numbers of input views. A parametric post-processing module detects braid regions and inpaints parameterized braid geometry into the generated latent map. Comparisons are reported against HAAR and HairStep on a 5% split of MultiHair using CLIP Score, Chamfer Distance, and point-cloud IoU, plus a user study reporting 63.8% preference on realistic inputs and 84.3% on stylized inputs.

Significance. If the central claims are supported, the paper makes a useful contribution: MultiHair addresses a real gap in diversity for 3D hair datasets, lineart conditioning is a sensible mechanism for suppressing lighting and texture variation while retaining strand topology, and the braid-specific parametric inpainting module targets a genuinely hard geometry problem. The qualitative demonstrations, including sketch-based editing and multi-view completion of occluded regions, are compelling. The paper also reports user studies and uses official pretrained baselines, which avoids an entire class of implementation-bias concerns. However, the evidence supporting the headline claim of 'arbitrary styles and viewpoints' is currently confined to the authors' own dataset and stylization pipeline; external validation on real photographs, hand-drawn sketches, or independent datasets would be needed before the generalization claim can be regarded as established.

major comments (3)
  1. [§5.2, Table 2] The quantitative comparison is computed on a 5% test split of the authors' own MultiHair dataset, which is roughly 23 hairstyles drawn from the same collection, rendered on the same discrete grid of 24 camera directions and 3 focal lengths, and stylized by the same ControlNet lineart-to-image pipeline used to create the conditioning signal. This setup does not yet support the paper's central claim of arbitrary styles and viewpoints. In addition, the baselines are official pretrained checkpoints that were not trained on MultiHair, so the comparison may systematically favor TANGLED regardless of method quality. The reported means over 10 stochastic samples are not accompanied by error bars, confidence intervals, or significance tests. I recommend adding an external evaluation (e.g., USC-HairSalon or Hair20k hairstyles, real portrait photos with unknown pose, and hand-drawn sketches), reporting per-style and per-viewpoint breakdowns, and providing variance or significance information for the main metrics.
  2. [§5.3] The paper states that 'the model trained on lineart outperforms the one trained on rendered images across all metrics,' but no numerical results for this ablation are reported anywhere. Since the argument that lineart filters out style noise is a load-bearing part of the abstract and §4.2, this ablation needs a concrete table with the same metrics used in Table 2, including the number of views and training setup. The row labeled Ours(R) in Table 2, described as 'trained on real images,' is also not explained: the source, size, and preprocessing of these real images should be specified, and it should be clarified whether Ours(R) is part of the same MultiHair test split.
  3. [§4.3, §5] The parametric braid inpainting module is presented as a central contribution, but it is evaluated only through qualitative figures and a global user study. There is no quantitative assessment of braid-specific outcomes, such as cycle-consistency of strand crossings, alignment with the braid mask, or comparison with and without the inpainting module. Given the paper's emphasis on complex braided styles and the claim that this module 'significantly reduces geometric distortions,' a controlled ablation with quantitative metrics is needed to support that claim.
minor comments (5)
  1. [Table 1] Several entries in Table 1 are incomplete or ambiguous: some rows show an asterisk or exclamation mark without a legend, and the 'Views' column is missing for some datasets. Please clean up the formatting and define all markers.
  2. [§3] The text says 'we generated 72 stylized images' for each hairstyle, while the preceding sentence describes 8 azimuths × 3 pitches × 3 focal lengths = 72 rendered images. It is unclear whether the 72 stylized images are additional to the 72 rendered images or whether stylization is applied only to a subset; please clarify the annotation counts.
  3. [§4.1, Eq. (2)] Equation (2) mixes an L2 distance, a cosine-similarity term, and an L2 distance without indicating relative weights or whether the norms are squared. Clarify the loss formulation.
  4. [§5.3] The final sentence of §5.3 trails off ('...in the image conditions.' appears to be an unfinished comparison). Please complete the sentence so the conclusion of the lineart-versus-rendered ablation is stated precisely.
  5. [References] Reference [21] appears to have a corrupted author list ('Chengan He Yi Zhou, Xin Sun'); please fix the citation to the Perm paper.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: TANGLED is an empirical generative pipeline whose claims rest on held-out evaluation, not on definitions, self-citations, or fitted-parameter renaming.

full rationale

TANGLED is an empirical image-to-3D generative model; there is no formal derivation chain whose conclusion is equivalent to its premises by construction. The conditioning path (multi-view lineart -> DINOv2 features -> cross-attention into a latent diffusion model -> strand decoder) is trained with an MSE denoising loss and evaluated on a held-out 5% split of the authors' MultiHair dataset. The reported improvements over HAAR and HairStep are empirical outcomes, not identities: the test inputs are lineart images derived from rendered hairstyles, while the outputs are predicted strand latents, and no fitted parameter is renamed as a prediction. The use of HAAR's strand-UV representation [56] is an external, independently published representation choice, not a self-citation, and the same-group citation [74] appears only in a related-work list and is not load-bearing. The fact that quantitative evaluation uses the authors' own dataset split is a legitimate external-validity limitation, but it does not make any result equivalent to its inputs by definition; therefore no circular step can be exhibited, and the circularity score is 0.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The paper is an empirical deep learning system, so there is no closed-form derivation. The free-parameter list includes only hand-adjustable braid geometry controls. The axioms are the standard probabilistic assumptions of latent diffusion plus domain assumptions that lineart extraction, segmentation, pose alignment, and the HAAR strand latent space behave as expected on stylized and braided hair. These assumptions are plausible but lightly validated in the paper.

free parameters (1)
  • Braid geometry controls (width, thickness, cross-sectional oscillation) = not reported, user-set
    In Sec. 4.3, braid geometry is generated with adjustable parameters along a guide strand; no values or ablations are provided, so these are hand-chosen controls.
assumptions (4)
  • standard math Latent diffusion denoising with DDIM sampling is a valid generative framework.
    Sec. 4.1 uses a standard diffusion training objective and DDIM at inference; correctness relies on established generative modeling assumptions.
  • domain assumption DINOv2 features extracted from lineart images are a sufficient conditioning signal for hair topology.
    Sec. 4.2 feeds concatenated DINOv2 lineart features through cross-attention, but no quantitative validation shows that these features preserve strand density and parting across styles.
  • domain assumption The off-the-shelf components (lineart detector, Grounding DINO, SAM, 3DDFAv2) work reliably on stylized and braided hair.
    Secs. 4.2 and 4.3 rely on these tools without reporting their failure rates or accuracy on the target domains.
  • domain assumption The strand UV latent space introduced by HAAR transfers to the MultiHair dataset and supports image-conditioned diffusion.
    Sec. 4.1 reuses HAAR's polyline latent representation and does not state whether the VAE is frozen or retrained on MultiHair.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TANGLED: Generating 3D Hair Strands from Images with Arbitrary Styles and Viewpoints." pith.science (2026). https://pith.science/paper/CIWDROMV

@misc{pith2026250206392,
  author       = {Pith},
  title        = {Pith review of: TANGLED: Generating 3D Hair Strands from Images with Arbitrary Styles and Viewpoints},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CIWDROMV}},
  note         = {Machine review of arXiv:2502.06392}
}
read the original abstract

Hairstyles are intricate and culturally significant with various geometries, textures, and structures. Existing text or image-guided generation methods fail to handle the richness and complexity of diverse styles. We present TANGLED, a novel approach for 3D hair strand generation that accommodates diverse image inputs across styles, viewpoints, and quantities of input views. TANGLED employs a three-step pipeline. First, our MultiHair Dataset provides 457 diverse hairstyles annotated with 74 attributes, emphasizing complex and culturally significant styles to improve model generalization. Second, we propose a diffusion framework conditioned on multi-view linearts that can capture topological cues (e.g., strand density and parting lines) while filtering out noise. By leveraging a latent diffusion model with cross-attention on lineart features, our method achieves flexible and robust 3D hair generation across diverse input conditions. Third, a parametric post-processing module enforces braid-specific constraints to maintain coherence in complex structures. This framework not only advances hairstyle realism and diversity but also enables culturally inclusive digital avatars and novel applications like sketch-based 3D strand editing for animation and augmented reality.

Figures

Figures reproduced from arXiv: 2502.06392 by the authors.

Figure 1
Figure 1. TANGLED brings creativity to life by generating high-quality 3D hairstyles from images of any style or viewpoint, seamlessly [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Dataset Annotation Process. Our annotation pipeline begins by processing rendered 3D hair strands with a line-art detector, and line-art sketches are combined with OpenPose [5] skeletal data for conditioning ControlNet. To enrich dataset diver￾sity, we further synthesize multi-view images, to cover variations in lighting, texture, and perspective. Finally, GPT-4 [43] generates detailed textual annotations for each h… view at source ↗
Figure 3
Figure 3. Architecture of our TANGLED . Our model takes hair images with arbitrary styles and viewpoints as conditions, and generate the 3D hair latent through the diffusion process. The conditions are randomly masked and cross-attention with the latent. At inference, we sample hair latent maps and feed the upsampled hair latent map to the strand decoder to extract the 3D hair strands. fidelity term is expressed as: Ldist = X… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4 [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Application showcase. Row 1 show the generated hairstyles from hand-drawn sketches. Row 2 illustrate hairstyle modifications(adding pigtails) by altering specific parts in the sketches from Row 1. Row 3-4 depict the process of generating outputs with braid using guidel…
Figure 6
Figure 6. Figure 6: Result gallery. TANGLED can generate realistic hairstyles from image conditions with various styles, including photographs, anime, and oil paintings. For more results, please refer to the supplementary video. Note that we manually specified the color for the generated …
Figure 7
Figure 7. Figure 7: Qualitative comparison. For various input images, our method produces more aligned and detailed hairstyles compared to HairStep and HAAR. For more comparison, please refer to [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 10
Figure 10. Figure 10: Generated results using single-view and multi￾view inputs. With single-view input(Row 1), the asymmetrical shoulder-length hair on the occluded side cannot be accurately re￾constructed, resulting in missing details. In contrast, multi-view inputs (Row 2) enable the mo…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

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

    cs.CV 2025-09 conditional novelty 6.0 of 10

    A synthetic-and-real trained transformer prior plus Gaussian-splatting optimization reconstructs strand-based 3D hairstyles from a single photograph.

  2. HairFormer: Transformer-Based Dynamic Neural Hair Simulation

    cs.GR 2025-07 conditional novelty 6.0 of 10

    A transformer-based two-stage network predicts static hair drapes and dynamic hair motion for arbitrary hairstyles and body poses in real time, trained with physics-inspired losses rather than pre-simulated data.

Reference graph

Works this paper leans on

79 extracted references · 74 canonical work pages · cited by 2 Pith papers

  1. [1]

    Dreamstudio: Ai-powered image gener- ation, 2025

    Stability AI. Dreamstudio: Ai-powered image gener- ation, 2025. Accessed: 2025-01-17. 2

    work page 2025

  2. [2]

    Artbreeder: Collaborative image creation using ai, 2025

    Artbreeder. Artbreeder: Collaborative image creation using ai, 2025. Accessed: 2025-01-17. 2

    work page 2025

  3. [3]

    Improving image generation with better captions

    James Betker, Gabriel Goh, Li Jing, Tim Brooks, Jianfeng Wang, Linjie Li, Long Ouyang, Juntang Zhuang, Joyce Lee, Yufei Guo, et al. Improving image generation with better captions. Computer Science. https://cdn. openai. com/papers/dall-e-3. pdf , 2(3):8,

    work page

  4. [4]

    Real-time hair rendering with hair meshes

    Gaurav Bhokare, Eisen Montalvo, Elie Diaz, and Cem Yuksel. Real-time hair rendering with hair meshes. In ACM SIGGRAPH 2024 Conference Papers , pages 1–10, 2024. 3

    work page 2024

  5. [5]

    Realtime multi-person 2d pose estimation us- ing part affinity fields

    Zhe Cao, Tomas Simon, Shih-En Wei, and Yaser Sheikh. Realtime multi-person 2d pose estimation us- ing part affinity fields. In Proceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 7291–7299, 2017. 4

    work page 2017

  6. [6]

    High-quality hair modeling from a single portrait photo

    Menglei Chai, Linjie Luo, Kalyan Sunkavalli, Nathan Carr, Sunil Hadap, and Kun Zhou. High-quality hair modeling from a single portrait photo. ACM Trans. Graph., 34(6):204–1, 2015. 3

    work page 2015

  7. [7]

    Autohair: Fully automatic hair modeling from a single image

    Menglei Chai, Tianjia Shao, Hongzhi Wu, Yanlin Weng, and Kun Zhou. Autohair: Fully automatic hair modeling from a single image. ACM Transactions on Graphics, 35(4), 2016. 3

    work page 2016

  8. [8]

    Dynamic hair manipulation in images and videos

    Menglei Chai, Lvdi Wang, Yanlin Weng, Xiaogang Jin, and Kun Zhou. Dynamic hair manipulation in images and videos. ACM Transactions on Graphics (TOG), 32(4):1–8, 2013. 3

    work page 2013

Show all 79 references
  1. [9]

    Single-view hair mod- eling for portrait manipulation

    Menglei Chai, Lvdi Wang, Yanlin Weng, Yizhou Yu, Baining Guo, and Kun Zhou. Single-view hair mod- eling for portrait manipulation. ACM Transactions on Graphics (TOG), 31(4):1–8, 2012. 3

    2012

  2. [10]

    A system of 3d hair style synthe- sis based on the wisp model

    Lieu-Hen Chen, Santi Saeyor, Hiroshi Dohi, and Mit- suru Ishizuka. A system of 3d hair style synthe- sis based on the wisp model. The Visual Computer, 15(4):159–170, 1999. 3

    1999

  3. [11]

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

    Matt Jen-Yuan Chiang, Benedikt Bitterli, Chuck Tap- pan, and Brent Burley. A practical and controllable hair and fur model for production path tracing. In ACM SIGGRAPH 2015 Talks, pages 1–1. 2015. 3

    2015

  4. [12]

    A statistical wisp model and pseudophysical approaches for inter- active hairstyle generation

    Byoungwon Choe and Hyeong-Seok Ko. A statistical wisp model and pseudophysical approaches for inter- active hairstyle generation. IEEE Transactions on Vi- sualization and Computer Graphics , 11(2):160–170,

  5. [13]

    Spin transformations of discrete surfaces

    Keenan Crane, Ulrich Pinkall, and Peter Schr ¨oder. Spin transformations of discrete surfaces. ACM Trans. Graph., 30, 2011. 6

    2011

  6. [14]

    Interactive hair simulation on the gpu using admm

    Gilles Daviet. Interactive hair simulation on the gpu using admm. In ACM SIGGRAPH 2023 Conference Proceedings, pages 1–11, 2023. 3

    2023

  7. [15]

    Objaverse: A universe of annotated 3d objects

    Matt Deitke, Dustin Schwenk, Jordi Salvador, Luca Weihs, Oscar Michel, Eli VanderBilt, Ludwig Schmidt, Kiana Ehsanit, Aniruddha Kembhavi, and Ali Farhadi. Objaverse: A universe of annotated 3d objects. In 2023 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (C...

    2023

  8. [16]

    Path tracing in produc- tion

    Luca Fascione, Johannes Hanika, Rob Piek ´e, Ryusuke Villemin, Christophe Hery, Manuel Gamito, Luke Emrose, and Andr´e Mazzone. Path tracing in produc- tion. In ACM SIGGRAPH 2018 Courses, pages 1–79

    2018

  9. [17]

    A multi-scale model for simulating liquid-hair interactions

    Yun Fei, Henrique Teles Maia, Christopher Batty, Changxi Zheng, and Eitan Grinspun. A multi-scale model for simulating liquid-hair interactions. ACM Transactions on Graphics (TOG) , 36(4):1–17, 2017. 3

    2017

  10. [18]

    Towards fast, accurate and stable 3d dense face alignment

    Jianzhu Guo, Xiangyu Zhu, Yang Yang, Fan Yang, Zhen Lei, and Stan Z Li. Towards fast, accurate and stable 3d dense face alignment. In European Confer- ence on Computer Vision , pages 152–168. Springer,

  11. [19]

    Ef- ficient diffusion training via min-snr weighting strat- egy

    Tiankai Hang, Shuyang Gu, Chen Li, Jianmin Bao, Dong Chen, Han Hu, Xin Geng, and Baining Guo. Ef- ficient diffusion training via min-snr weighting strat- egy. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 7441–7451,

  12. [20]

    Digital salon: An ai and physics-driven tool for 3d hair grooming and simula- tion

    Chengan He, Jorge Alejandro Amador Herrera, Yi Zhou, Zhixin Shu, Xin Sun, Yao Feng, S ¨oren Pirk, Dominik L Michels, Meng Zhang, Tuanfeng Y Wang, and Holly Rushmeier. Digital salon: An ai and physics-driven tool for 3d hair grooming and simula- tion. ACM SIGGRAPH Asia 2024 Rea...

    2024

  13. [21]

    Perm: A parametric representa- tion for multi-style 3d hair modeling

    Chengan He, Xin Sun, Zhixin Shu, Fujun Luan, S ¨oren Pirk, Jorge Alejandro Amador Herrera, Dominik L Michels, Tuanfeng Y Wang, Meng Zhang, Holly Rushmeier, et al. Perm: A parametric representa- tion for multi-style 3d hair modeling. arXiv preprint arXiv:2407.19451, 2024. 2, 3

    2024 arXiv

  14. [22]

    Augmented mass-spring model for real-time dense hair simulation

    Jorge Alejandro Amador Herrera, Yi Zhou, Xin Sun, Zhixin Shu, Chengan He, S ¨oren Pirk, and Do- minik L Michels. Augmented mass-spring model for real-time dense hair simulation. arXiv preprint arXiv:2412.17144, 2024. 3

    2024

  15. [23]

    Classifier-free dif- fusion guidance

    Jonathan Ho and Tim Salimans. Classifier-free dif- fusion guidance. arXiv preprint arXiv:2207.12598 ,

    arXiv

  16. [24]

    Sag-free initialization for strand-based hybrid hair simulation

    Jerry Hsu, Tongtong Wang, Zherong Pan, Xifeng Gao, Cem Yuksel, and Kui Wu. Sag-free initialization for strand-based hybrid hair simulation. ACM Transac- tions on Graphics (TOG), 42(4):1–14, 2023. 3

    2023

  17. [25]

    Real-time physically guided hair interpolation

    Jerry Hsu, Tongtong Wang, Zherong Pan, Xifeng Gao, Cem Yuksel, and Kui Wu. Real-time physically guided hair interpolation. ACM Trans. Graph., 43(4), July 2024. 6

    2024

  18. [27]

    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 Transactions on Graphics (ToG) , 34(4):1–9,

  19. [28]

    Capturing braided hairstyles

    Liwen Hu, Chongyang Ma, Linjie Luo, Li-Yi Wei, and Hao Li. Capturing braided hairstyles. ACM Transac- tions on Graphics (TOG), 33(6):1–9, 2014. 6

    2014

  20. [29]

    Towards realtime: A hybrid physics-based method for hair animation on gpu

    Li Huang, Fan Yang, Chendi Wei, Yu Ju (Edwin) Chen, Chun Yuan, and Ming Gao. Towards realtime: A hybrid physics-based method for hair animation on gpu. 6(3), Aug. 2023. 6

    2023

  21. [30]

    Interactive mul- tiresolution hair modeling and editing

    Tae-Yong Kim and Ulrich Neumann. Interactive mul- tiresolution hair modeling and editing. ACM Transac- tions on Graphics (TOG), 21(3):620–629, 2002. 3

    2002

  22. [31]

    Segment anything

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alexander C Berg, Wan-Yen Lo, et al. Segment anything. In Proceedings of the IEEE/CVF International Conference on Computer Vi- sion, pages 4015–4026, 2023. 5, 6

    2023

  23. [32]

    Real-time an- imation of human hair modeled in strips

    Chuan Koon Koh and Zhiyong Huang. Real-time an- imation of human hair modeled in strips. In Com- puter Animation and Simulation 2000: Proceedings of the Eurographics Workshop in Interlaken, Switzer- land, August 21–22, 2000 , pages 101–110. Springer,

  24. [33]

    Generation of 3d hair model from multiple images

    WM Kong. Generation of 3d hair model from multiple images. The Journal of the Institute of Image Informa- tion and Television Engineers , 52(9):109–114, 1998. 3

    1998

  25. [34]

    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 Confer- ence Papers, pages 1–8, 2022. 2, 3

    2022

  26. [35]

    Blip: Bootstrapping language-image pre-training for unified vision-language understanding and genera- tion

    Junnan Li, Dongxu Li, Caiming Xiong, and Steven Hoi. Blip: Bootstrapping language-image pre-training for unified vision-language understanding and genera- tion. In International conference on machine learning, pages 12888–12900. PMLR, 2022. 7

    2022

  27. [36]

    An enhanced framework for real-time hair animation

    Wenqi Liang and Zhiyong Huang. An enhanced framework for real-time hair animation. In 11th Pa- cific Conference onComputer Graphics and Appli- cations, 2003. Proceedings. , pages 467–471. IEEE,

    2003

  28. [37]

    Grounding dino: Marry- ing dino with grounded pre-training for open-set ob- ject detection

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Qing Jiang, Chunyuan Li, Jian- wei Yang, Hang Su, et al. Grounding dino: Marry- ing dino with grounded pre-training for open-set ob- ject detection. In European Conference on Computer Vision, pages 38–55. Sp...

    2025

  29. [38]

    Decoupled weight decay regularization

    I Loshchilov. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101, 2017. 7

    2017 arXiv

  30. [39]

    Re- paint: Inpainting using denoising diffusion probabilis- tic models

    Andreas Lugmayr, Martin Danelljan, Andres Romero, Fisher Yu, Radu Timofte, and Luc Van Gool. Re- paint: Inpainting using denoising diffusion probabilis- tic models. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 11461–11471, 2022. 6

    2022

  31. [40]

    Hairmony: Fairness-aware hairstyle classification

    Givi Meishvili, James Clemoes, Charlie Hewitt, Zafi- irah Hosenie, Xiao Xian, Martin de La Gorce, Tibor Takacs, Tadas Baltruˇsaitis, Antonio Criminisi, Chyna McRae, Nina Jablonski, and Marta Wilczkowiak. Hairmony: Fairness-aware hairstyle classification. In SIGGRAPH Asia 2024 ...

    2024

  32. [41]

    Strand-accurate multi-view hair capture

    Giljoo Nam, Chenglei Wu, Min H Kim, and Yaser Sheikh. Strand-accurate multi-view hair capture. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pages 155–164,

  33. [42]

    Modelling and animat- ing cartoon hair with nurbs surfaces

    Paul Noble and Wen Tang. Modelling and animat- ing cartoon hair with nurbs surfaces. In Proceedings Computer Graphics International, 2004. , pages 60–

    2004

  34. [43]

    Chatgpt: A large language model for general-purpose tasks, 2025

    OpenAI. Chatgpt: A large language model for general-purpose tasks, 2025. Accessed: 2025-01-17. 4

    2025

  35. [44]

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

    2024

  36. [45]

    Capture of hair geometry from multiple images

    Sylvain Paris, Hector M Briceno, and Franc ¸ois X Sil- lion. Capture of hair geometry from multiple images. ACM transactions on graphics (TOG), 23(3):712–719,

  37. [46]

    Modelling and rendering techniques for african hairstyles

    Deborah Patrick, Shaun Bangay, and Adele Lobb. Modelling and rendering techniques for african hairstyles. In Proceedings of the 3rd international conference on Computer graphics, virtual reality, vi- sualisation and interaction in Africa , pages 115–124,

  38. [47]

    Generalized helicoids for modeling hair geometry

    Emmanuel Piuze, Paul G Kry, and Kaleem Siddiqi. Generalized helicoids for modeling hair geometry. In Computer Graphics Forum , volume 30, pages 247–

  39. [48]

    Dreamfusion: Text-to-3d using 2d dif- fusion

    Ben Poole, Ajay Jain, Jonathan T Barron, and Ben Mildenhall. Dreamfusion: Text-to-3d using 2d dif- fusion. arXiv e-prints, pages arXiv–2209, 2022. 3

    2022

  40. [49]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sas- try, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning, ...

    2021

  41. [50]

    High- resolution image synthesis with latent diffusion mod- els

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High- resolution image synthesis with latent diffusion mod- els. In 2022 IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition (CVPR) , pages 10674– 10685, 2022. 2

    2022

  42. [51]

    U-net: Convolutional networks for biomedical im- age segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical im- age segmentation. In Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, par...

    2015

  43. [52]

    3d hair synthe- sis using volumetric variational autoencoders

    Shunsuke Saito, Liwen Hu, Chongyang Ma, Hikaru Ibayashi, Linjie Luo, and Hao Li. 3d hair synthe- sis using volumetric variational autoencoders. ACM Transactions on Graphics (TOG) , 37(6):1–12, 2018. 3

    2018

  44. [53]

    Ct2hair: High-fidelity 3d hair mod- eling using computed tomography

    Yuefan Shen, Shunsuke Saito, Ziyan Wang, Olivier Maury, Chenglei Wu, Jessica Hodgins, Youyi Zheng, and Giljoo Nam. Ct2hair: High-fidelity 3d hair mod- eling using computed tomography. ACM Transactions on Graphics (TOG), 42(4):1–13, 2023. 2, 3

    2023

  45. [54]

    Deepsketchhair: Deep sketch-based 3d hair modeling

    Yuefan Shen, Changgeng Zhang, Hongbo Fu, Kun Zhou, and Youyi Zheng. Deepsketchhair: Deep sketch-based 3d hair modeling. IEEE transactions on visualization and computer graphics , 27(7):3250– 3263, 2020. 3

    2020

  46. [55]

    Neural haircut: Prior-guided strand-based hair recon- struction

    Vanessa Sklyarova, Jenya Chelishev, Andreea Dogaru, Igor Medvedev, Victor Lempitsky, and Egor Zakharov. Neural haircut: Prior-guided strand-based hair recon- struction. In Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision, pages 19762– 19773, 2023. 3, 4

    2023

  47. [56]

    Black, and Justus Thies

    Vanessa Sklyarova, Egor Zakharov, Otmar Hilliges, Michael J. Black, and Justus Thies. Text-conditioned generative model of 3d strand-based human hairstyles. In IEEE/CVF Conf. on Computer Vision and Pattern Recognition (CVPR), June 2024. 2, 3, 4, 7

    2024

  48. [57]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502, 2020. 7

    2010 arXiv

  49. [58]

    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 Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Reco...

    2024

  50. [59]

    Neus: Learning neural implicit surfaces by volume render- ing for multi-view reconstruction

    Peng Wang, Lingjie Liu, Yuan Liu, Christian Theobalt, Taku Komura, and Wenping Wang. Neus: Learning neural implicit surfaces by volume render- ing for multi-view reconstruction. Advances in Neu- ral Information Processing Systems, 34:27171–27183,

  51. [60]

    Hair design based on the hierarchical cluster hair model

    Tao Wang and Xue Dong Yang. Hair design based on the hierarchical cluster hair model. Geometric mod- eling: techniques, applications, systems and tools , pages 329–359, 2004. 3

    2004

  52. [61]

    Adaptive shells for efficient neural radiance field rendering

    Zian Wang, Tianchang Shen, Merlin Nimier-David, Nicholas Sharp, Jun Gao, Alexander Keller, Sanja Fi- dler, Thomas M¨uller, and Zan Gojcic. Adaptive shells for efficient neural radiance field rendering. arXiv preprint arXiv:2311.10091, 2023. 3

    2023 arXiv

  53. [62]

    A trigonal prism-based method for hair image generation

    Yasuhiko Watanabe and Yasuhito Suenaga. A trigonal prism-based method for hair image generation. IEEE Computer Graphics and applications , 12(01):47–53,

  54. [63]

    Monohair: High-fidelity hair model- ing 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 model- ing from a monocular video. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 24164–2417...

    2024

  55. [64]

    Neuralhdhair: Automatic high-fidelity hair modeling from a single image using implicit neural representations

    Keyu Wu, Yifan Ye, Lingchen Yang, Hongbo Fu, Kun Zhou, and Youyi Zheng. Neuralhdhair: Automatic high-fidelity hair modeling from a single image using implicit neural representations. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages ...

    2022

  56. [65]

    Om- niobject3d: Large-vocabulary 3d object dataset for re- alistic perception, reconstruction and generation

    Tong Wu, Jiarui Zhang, Xiao Fu, Yuxin Wang, Ji- awei Ren, Liang Pan, Wayne Wu, Lei Yang, Jiaqi Wang, Chen Qian, Dahua Lin, and Ziwei Liu. Om- niobject3d: Large-vocabulary 3d object dataset for re- alistic perception, reconstruction and generation. In 2023 IEEE/CVF Conference o...

    2023

  57. [66]

    Hairbrush for immersive data-driven hair modeling

    Jun Xing, Koki Nagano, Weikai Chen, Haotian Xu, Li-yi Wei, Yajie Zhao, Jingwan Lu, Byungmoon Kim, and Hao Li. Hairbrush for immersive data-driven hair modeling. In Proceedings of the 32Nd Annual ACM Symposium on User Interface Software and Technol- ogy, pages 263–279, 2019. 3

    2019

  58. [67]

    V-hairstudio: an inter- active tool for hair design

    Zhan Xu and Xue Dong Yang. V-hairstudio: an inter- active tool for hair design. IEEE Computer Graphics and Applications, 21(3):36–43, 2001. 3

    2001

  59. [68]

    The cluster hair model

    Xue Dong Yang, Zhan Xu, Jun Yang, and Tao Wang. The cluster hair model. Graphical Models, 62(2):85– 103, 2000. 3

    2000

  60. [69]

    Hair20k: A large 3d hairstyle database for hair modeling

    Chengan He Yi Zhou, Xin Sun. Hair20k: A large 3d hairstyle database for hair modeling. 2024. 2, 3

    2024

  61. [70]

    Hair meshes

    Cem Yuksel, Scott Schaefer, and John Keyser. Hair meshes. ACM Transactions on Graphics (TOG) , 28(5):1–7, 2009. 3

    2009

  62. [71]

    Hu- man hair reconstruction with strand-aligned 3d gaus- sians

    Egor Zakharov, Vanessa Sklyarova, Michael Black, Giljoo Nam, Justus Thies, and Otmar Hilliges. Hu- man hair reconstruction with strand-aligned 3d gaus- sians. In European Conference on Computer Vision , pages 409–425. Springer, 2025. 3

    2025

  63. [72]

    Hairdiffu- sion: Vivid multi-colored hair editing via latent diffu- sion

    Yu Zeng, Yang Zhang, Jiachen Liu, Linlin Shen, Kai- jun Deng, Weizhao He, and Jinbao Wang. Hairdiffu- sion: Vivid multi-colored hair editing via latent diffu- sion. arXiv e-prints, pages arXiv–2410, 2024. 2

    2024

  64. [73]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision , pages 3836– 3847, 2023. 4, 5

    2023

  65. [74]

    Clay: A controllable large-scale gener- ative model for creating high-quality 3d assets

    Longwen Zhang, Ziyu Wang, Qixuan Zhang, Qiwei Qiu, Anqi Pang, Haoran Jiang, Wei Yang, Lan Xu, and Jingyi Yu. Clay: A controllable large-scale gener- ative model for creating high-quality 3d assets. ACM Transactions on Graphics (TOG), 43(4):1–20, 2024. 2

    2024

  66. [75]

    Hair-gan: Recovering 3d hair structure from a single image using generative adversarial networks

    Meng Zhang and Youyi Zheng. Hair-gan: Recovering 3d hair structure from a single image using generative adversarial networks. Visual Informatics, 3(2):102– 112, 2019. 3

    2019

  67. [76]

    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 Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogn...

  68. [77]

    Groomgen: A high- quality generative hair model using hierarchical la- tent representations

    Yuxiao Zhou, Menglei Chai, Alessandro Pepe, Markus Gross, and Thabo Beeler. Groomgen: A high- quality generative hair model using hierarchical la- tent representations. ACM Transactions on Graphics (TOG), 42(6):1–16, 2023. 3

    2023

  69. [78]

    Groomcap: High-fidelity prior-free hair capture

    Yuxiao Zhou, Menglei Chai, Daoye Wang, Sebastian Winberg, Erroll Wood, Kripasindhu Sarkar, Markus Gross, and Thabo Beeler. Groomcap: High-fidelity prior-free hair capture. ACM Transactions on Graph- ics (TOG), 43(6):1–15, 2024. 3

    2024

  70. [79]

    Hairnet: Single- view hair reconstruction using convolutional neural networks

    Yi Zhou, Liwen Hu, Jun Xing, Weikai Chen, Han- Wei Kung, Xin Tong, and Hao Li. Hairnet: Single- view hair reconstruction using convolutional neural networks. In Proceedings of the European Conference on Computer Vision (ECCV), pages 235–251, 2018. 2, 3

    2018

  71. [256]

    Wiley Online Library, 2011. 3

    2011

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.