Pith. sign in

REVIEW 4 major objections 4 minor 41 references

StyleDiT: A Unified Framework for Diverse Child and Partner Faces Synthesis with Style Latent Diffusion Transformer

T0 review · 4 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper proposes StyleDiT, a unified framework that generates diverse child faces from two parents and, for the first time, predicts a partner's face from a child and one parent.

desk verdict Architecturally inventive, but the synthetic training loop makes the kinship claims hard to trust. read the letter →

arxiv 2412.10785 v1 pith:OAWQYU7O submitted 2024-12-14 cs.CV

Pin-Yen Chiu , Dai-Jie Wu , Po-Hsun Chu , Chia-Hsuan Hsu , Hsiang-Chen Chiu , Chih-Yu Wang , Jun-Cheng Chen This is my paper · ORCID
classification cs.CV
keywords kinshipfacesynthesispartnerpredictionstylelatentdiffusionGANspaceclassifier-freeguidanceageandgendercontroldiversity-fidelitytrade-offtransformer
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

The paper tries to establish that kinship face synthesis can be unified into a single generative framework that is both diverse and faithful: sampling a StyleGAN style latent with a transformer-based diffusion model yields many plausible child faces for one parent pair, with fine-grained control over age and gender, instead of the single deterministic output of earlier methods. It further claims to be the first to handle partner prediction, generating the missing parent from a child's face and one parent's face in the same framework. The practical stake is that one model, trained once on simulated family triplets, could serve applications from finding long-lost family members to reconstructing a missing person's appearance, while letting a user tune how much the output resembles each input face.

What carries the argument

The central object is the StyleDiT denoising transformer operating in StyleGAN2's $S$ space, the 9088-dimensional style-parameter space. Style parameters are split into 26 groups, each projected to a 512-dimensional token; a transformer denoises a noisy token set conditioned on two encoded input style latents via cross-attention, and an untokenizer maps tokens back to $S$. Relational Trait Guidance (RTG) is the control mechanism: it treats each conditioning latent with its own classifier-free guidance scale, so the user can strengthen or weaken each parent's influence independently and trade diversity against fidelity.

What would settle it

Encode real parent-child triples from FIW, TSKinFace, or FF-Database into S space and check whether each child latent lies near $(S_F+S_M)/2$; separately, generate predicted children from real parents and compare identity similarity to real siblings versus age-, gender-, and ethnicity-matched unrelated adults. If real child latents deviate systematically from the midpoint or predicted children resemble unrelated matches as much as real siblings, the central claim is refuted.

Watch

Extended reading notes

Core claim

The central claim is that the distribution of kinship relationships can be modeled in StyleGAN2's style-latent space by a conditional denoising transformer, and that a multi-way extension of classifier-free guidance gives independent control over each conditioning face. For child prediction the model takes two encoded parent latents and produces a distribution of child latents; for partner prediction it takes a child latent and one parent latent and produces the other parent. The paper demonstrates the partner task is not solvable by the naive linear rearrangement $S_{M/F} = 2S_C - S_{F/M}$ of the child-blending rule $S_C = \frac{1}{2}(S_F + S_M)$, arguing that latent extrapolation falls outside the trained distribution and a learned diffusion prior is required instead.

Load-bearing premise

The paper builds its 100,000 training triplets on the assumption that a child's style latent is the midpoint of the parents' style latents; if real children are not linear blends of their parents, the model learns a synthetic relationship and the reported resemblance to real children may reflect age, gender, and ethnicity rather than true kinship.

Editorial extensions

If this is right

  • For a fixed parent pair, a user can dial the generated child's resemblance toward the father or the mother by adjusting that parent's guidance scale.
  • The same trained framework performs both child prediction and partner prediction, so a single model covers both tasks.
  • Sampling different noise instances under the same conditions yields many plausible children, overcoming the single-output limitation of deterministic methods.
  • Fine-grained age (0-99) and gender control is available at inference time through the attribute-conditioned encoder, without retraining per attribute.
  • In the user study, StyleDiT received the most first-place votes among the four methods for child prediction.

Reading between the lines

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

  • If the linear-blend training assumption is the real source of the model's behavior, then a direct consequence the authors do not explore is that the framework is only as kinship-realistic as the synthetic prior; testing on families with multiple biological siblings would separate learned kinship from a learned 'average face' prior.
  • The per-condition guidance scales suggest a general recipe for multi-input latent diffusion: treat each conditioning image as an independent guidance direction, which could transfer to other two-image tasks such as parent-similarity editing, style mixing, or identity-preserving portrait interpolation.
  • Partner prediction as latent extrapolation is a distinct claim: the diffusion prior regularizes an out-of-distribution extrapolation that linear algebra fails at. An interesting test is whether the same architecture also extrapolates other latent relationships, such as grandparent faces or age-progressed children, or simply collapses to the mean of the conditioning pair.
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

4 major / 4 minor

Summary. The paper proposes StyleDiT, a framework that combines a StyleGAN2 decoder with a transformer-based diffusion model operating in StyleGAN's S space to synthesize child faces from two parent images and partner faces from a child image and one parent image. A Relational Trait Guidance (RTG) mechanism is introduced to allow independent control over each conditioning image and to trade off diversity against fidelity. The model is trained on 100,000 synthetic triplets generated by linear interpolation of parent style latents (Eq. 10). Evaluations are conducted on FIW, TSKinFace, and FF-Database using ArcFace identity similarity and a pairwise cosine-similarity diversity score, together with a 100-participant user study. The paper claims to achieve an excellent balance between diversity and fidelity and to be the first to address partner face prediction.

Significance. If the claims are supported, the paper would make a useful contribution: the unified child/partner synthesis task is novel, the RTG mechanism is a clean extension of classifier-free guidance, and the qualitative results show high-quality and diverse outputs. The paper also includes a large user study and ablations. However, the central evidence is undercut by a circularity between the synthetic training distribution and the evaluation protocol. The model is trained to invert the exact linear relationship used to generate its training labels, so the identity-similarity scores on real kinship test sets largely measure how well real families conform to that synthetic prior rather than whether the model captures genuine kinship. The authors themselves acknowledge in Appendix G.1 that real kinship relationships cannot be fully captured by simple linear interpolation. The absence of error bars, significance tests, code, and data further limits verification. The architectural ideas are interesting, but the current results do not establish that StyleDiT learns a kinship function beyond the trivial linear blending.

major comments (4)
  1. [Section 4.1, Appendix D.1 Eq. (10)] The synthetic training set is generated by S_C = 1/2(S_F + S_M) in StyleGAN's S space, and the diffusion model is trained to predict exactly this conditional relationship. The main results in Table 1 evaluate identity similarity on real kinship test sets, so the reported ID Sim values (0.7003, 0.7244, 0.7124) measure how well real triplets conform to the paper's linear-blending prior, not whether StyleDiT captures real kinship. This is a load-bearing issue because the authors state in Appendix G.1 that real kinship 'cannot be fully captured by simple linear interpolation.' To support the claim of outperforming prior methods, the paper must include a direct linear-interpolation baseline using the same image encoder and StyleGAN decoder under the identical evaluation protocol; if that baseline achieves comparable ID Sim, the method's advantage over the trivial construction is not established.
  2. [Section 4.3, Tables 1–5] All quantitative results are reported as point estimates without error bars, confidence intervals, or significance tests. The differences between methods are small (e.g., FIW ID Sim of 0.7003 for StyleDiT versus 0.7132 for StyleGene), and the user study in Table 3 reports average ranks without statistical testing. The paper should provide repeated-run statistics (e.g., multiple random seeds and dataset splits) and paired significance tests to substantiate the central comparative claim that StyleDiT strikes a better fidelity-diversity balance than prior methods.
  3. [Appendix D.1 Eq. (11), Appendix D.2 Table 3] Partner prediction is claimed as a novel contribution, but the training data for this task is derived by rearranging the same linear relation (Eq. 11: S_M/F = 2*S_C - S_F/M), and Table 3 evaluates against ground-truth partners with no baseline and no error bars. The paper states that linear extrapolation fails to produce optimal results but provides only a qualitative example (Appendix D.1, Fig. 4). A quantitative comparison against the linear extrapolation baseline with the same encoder and decoder is necessary to demonstrate that the learning-based approach adds value beyond the algebraic rearrangement of Eq. (10).
  4. [Section 4.4, Appendix G] The ablation showing that fine-tuning or training solely on real data reduces ID Sim (Table 4) is interpreted as evidence of real-data quality issues. An alternative explanation is that the evaluation metric rewards the synthetic prior: a model trained on synthetic linear-interpolation data will score highly on real data that happens to approximate the linear rule. Without a real-trained model evaluated on a task that does not presuppose the linear relation (e.g., kinship verification or a human-perception study with proper controls), the ablation does not resolve which explanation is correct, so the paper's dismissal of real-data training is not fully supported.
minor comments (4)
  1. [Appendix H] The limitation section only discusses GAN inversion accuracy; it omits the more central limitation that the model is trained entirely on synthetic data generated by the linear-interpolation assumption, which is acknowledged elsewhere in the paper as an incomplete model of real kinship. This should be stated explicitly in the limitations discussion.
  2. [Section 4.3, Eq. (4)] The diversity score is defined as a mean cosine similarity, but the notation xi·xj/||xi||||xj|| is redundant given the explicit norm terms. More importantly, the metric is sensitive to image quality and does not control for attributes such as age or gender, so lower-quality outputs (e.g., ChildPredictor) may appear artificially more diverse; this should be discussed as a limitation.
  3. [Throughout] There are several typos and formatting inconsistencies: reference [25] misspells 'European', the figure legend in Fig. 6 writes 'StyleDit' instead of 'StyleDiT', and Table 3's '1,000 votes' should clarify that it aggregates 100 participants across 10 questions per participant. The paper would also benefit from stating whether the reported ID Sim values are averaged over the 20 generated samples per family.
  4. [Appendix F] The user study does not report inter-rater reliability or a statistical comparison between methods, and the sample selection of only 7-10 parent sets per session may limit representativeness. These details matter for interpreting the average ranks and first-place votes in Table 3.

Circularity Check

2 steps flagged · score 7.0 of 10

Child and partner 'predictions' are learned inverses of the paper's own linear equations S_C=(S_F+S_M)/2 and S_M/F=2S_C−S_F/M.

  1. self definitional [Section 4.1 (Datasets), Section 3.3 (training objective), Appendix D.1 Eq. (10)]
    "(Appendix D.1) 'a child representation, denoted as SC, can be generated through a linear combination of the style latent codes of the father, SF, and the mother, SM, using an equal weighting (half scale) for both in the style latent space... SC = 1/2(SF + SM).' (Section 3.3) 'We train the network by MSE loss to compare the denoised style latent S∗ against the original style latent S.'"

    The training target for child prediction is not measured from real kinship; it is constructed by Eq. (10) as the average of the two conditioning latents. Because the diffusion transformer is optimized by MSE to output exactly that target, the learned child latent is, at inference, a stochastic linear blend of the input parents' latents. The paper then presents this as modeling a complex kinship distribution, but the model has only inverted its own data-generation equation. Real-data identity-similarity scores therefore measure how well real families conform to the paper's linear-mixing rule, not an independently learned law of facial inheritance. Appendix G.1 admits real kinship 'cannot be fully captured by simple linear interpolation between the style latents of parents.'

  2. self definitional [Section 4.1 (Datasets), Appendix D.1 Eq. (11), Section 3.3]
    "(Appendix D.1) 'Additionally, to generate a parent’s representation from the child and the other parent’s latents, we could analogically rearrange the equation: SM/F = (2 × SC − SF/M ).' (Section 4.1) 'This simulated dataset is used for training both child and partner face prediction.'"

    Partner targets are generated by Eq. (11), an algebraic rearrangement of the child-generation equation, from the same conditioning images used as network inputs. Training the model to predict these targets makes 'partner prediction' a learned implementation of 2*S_child − S_parent. There is no external measurement of partner appearance used to define the learning signal; the framework's claimed capability is thus forced by the paper's own linear construction. The absence of any baseline for partner synthesis (Section 4.1: 'we found no existing studies formally addressing this problem') leaves this task as a self-defined inversion rather than an empirically grounded prediction.

full rationale

The main derivation chain is internally consistent but not independent. Section 4.1 builds a 100k-pair simulated dataset whose child latents are generated by linear interpolation between the parental style latents, and Appendix D.1 states the exact rule S_C = 1/2(S_F+S_M) and its partner re-arrangement S_M/F = 2S_C − S_F/M. The StyleDiT denoiser is trained with MSE against these generated latents, so the model's outputs are, by construction, samples from the paper's own generative equations. Real-data evaluations on FIW, TSKinFace, and FF-Database are external and provide some empirical check, but they cannot break the circularity: the model can only emit latents in the linear span of the inputs, so the reported identity-similarity numbers reflect how well real families obey the synthetic linear rule. The paper itself concedes this rule is insufficient for real data (Appendix G.1). The self-citation to KinStyle [4] for the linear-interpolation premise is noted but is not the decisive circular step; the decisive step is the training-target construction. Because the central child and partner generation capabilities reduce to learned inverses of Eqs. (10)-(11), a score of 7 is warranted. No evidence of renaming-known-result or imported-uniqueness-theorem circularity was found.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim rests on a chain of domain assumptions, most importantly that kinship is linear interpolation in StyleGAN latent space and that random face pairs can stand in for parents. These assumptions are not derived in the paper; they are borrowed from prior work and medical citations. The method itself adds a learned conditional diffusion model and a guidance mechanism, but the ground truth it learns is the synthetic interpolation process.

free parameters (3)
  • RTG guidance scales g_1, g_2 = 1.2 for child prediction; 1.2 and 0.0 for partner prediction
    Hand-selected inference hyperparameters controlling how strongly each input condition influences the output; central to the claimed diversity-fidelity trade-off.
  • interpolation weight w for synthetic child latents = not reported
    Used with variable weights to generate 100,000 synthetic child latents from parent latents; the distribution of w determines the learned conditional distribution and is never specified.
  • learned null-condition embedding p = learned, R^{1x18x512}
    Trained null condition for RTG (Appendix B.2); its initialization and training schedule are not given.
assumptions (5)
  • domain assumption A child's facial appearance is well approximated by linear interpolation of parental StyleGAN style latents.
    Used to create all training triplets (Section 4.1, Appendix D.1 Eq. 10). If false, the model learns a synthetic relationship rather than real kinship.
  • domain assumption Random male-female face pairs (CelebA-HQ) and random StyleGAN samples are valid stand-ins for real parents.
    70k random pairs and 30k sampled faces form the training parents; no real spousal relationship or shared environment is represented.
  • domain assumption ArcFace cosine similarity is a valid proxy for kinship identity similarity.
    Used for DS and ID Sim metrics; no calibration against human perception or kinship-verification benchmarks is provided.
  • domain assumption StyleSpace S space provides disentangled, editable facial attributes.
    Inherited from StyleSpace [36] and used for tokenization and attribute control.
  • standard math Standard diffusion and DDIM theory is applicable to the 9088-dimensional style latent.
    Background assumption, not in dispute.

how reviews work

0 comments
Cite this review

Pith. "Pith review of StyleDiT: A Unified Framework for Diverse Child and Partner Faces Synthesis with Style Latent Diffusion Transformer." pith.science (2026). https://pith.science/paper/OAWQYU7O

@misc{pith2026241210785,
  author       = {Pith},
  title        = {Pith review of: StyleDiT: A Unified Framework for Diverse Child and Partner Faces Synthesis with Style Latent Diffusion Transformer},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OAWQYU7O}},
  note         = {Machine review of arXiv:2412.10785}
}
read the original abstract

Kinship face synthesis is a challenging problem due to the scarcity and low quality of the available kinship data. Existing methods often struggle to generate descendants with both high diversity and fidelity while precisely controlling facial attributes such as age and gender. To address these issues, we propose the Style Latent Diffusion Transformer (StyleDiT), a novel framework that integrates the strengths of StyleGAN with the diffusion model to generate high-quality and diverse kinship faces. In this framework, the rich facial priors of StyleGAN enable fine-grained attribute control, while our conditional diffusion model is used to sample a StyleGAN latent aligned with the kinship relationship of conditioning images by utilizing the advantage of modeling complex kinship relationship distribution. StyleGAN then handles latent decoding for final face generation. Additionally, we introduce the Relational Trait Guidance (RTG) mechanism, enabling independent control of influencing conditions, such as each parent's facial image. RTG also enables a fine-grained adjustment between the diversity and fidelity in synthesized faces. Furthermore, we extend the application to an unexplored domain: predicting a partner's facial images using a child's image and one parent's image within the same framework. Extensive experiments demonstrate that our StyleDiT outperforms existing methods by striking an excellent balance between generating diverse and high-fidelity kinship faces.

Figures

Figures reproduced from arXiv: 2412.10785 by the authors.

Figure 1
Figure 1. StyleDiT showcases dual capabilities: it synthesizes child faces from parents’ images and generates partner faces from a child’s [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overview of the proposed framework. For both child and partner prediction tasks, input images are first encoded using [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The effect of different guidance scales during inference. Progressing from left to right, each image results from varying pairs of [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: This figure compares the facial diversity in results of [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: The qualitative comparison of synthesized children’s faces by StyleDiT and baselines. The first three columns in both (a) FIW [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: The results demonstrate our framework can synthesize [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 1
Figure 1. Figure 1: The overview of our image encoder. The input image is first encoded into latent code using e4e. Then, the latent code is adjusted [PITH_FULL_IMAGE:figures/full_fig_p012_1.png]
Figure 2
Figure 2. Figure 2: The comparison of controlling age, gender, and frontalization while preserving skin tone. For the first and second rows, our [PITH_FULL_IMAGE:figures/full_fig_p012_2.png]
Figure 3
Figure 3. Figure 3: The visual comparison showcases images from the FIW dataset, the FIW dataset after applying super-resolution [ [PITH_FULL_IMAGE:figures/full_fig_p014_3.png]
Figure 4
Figure 4. Figure 4: The visual comparison of partner face synthesis. [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: Our framework allows for fine-grained age and gender control, ranging from 0 to 99 years, offering flexibility to generate faces [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: Example questions in our user study. (a) An example question in our survey assessing the resemblance of synthesized child faces [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: Raw result that we collect from the user study. (a) [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 34 canonical work pages

  1. [1]

    Im- age2stylegan: How to embed images into the stylegan la- tent space? In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019

    Rameen Abdal, Yipeng Qin, and Peter Wonka. Im- age2stylegan: How to embed images into the stylegan la- tent space? In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019. 2

    work page 2019

  2. [2]

    Differential facial resemblance of young children to their parents: who do children look like more? Evolution and Human behavior, 28(2):135–144, 2007

    Alexandra Alvergne, Charlotte Faurie, and Michel Ray- mond. Differential facial resemblance of young children to their parents: who do children look like more? Evolution and Human behavior, 28(2):135–144, 2007. 2, 5, 7

    work page 2007

  3. [3]

    Spatext: Spatio-textual representation for con- trollable image generation

    Omri Avrahami, Thomas Hayes, Oran Gafni, Sonal Gupta, Yaniv Taigman, Devi Parikh, Dani Lischinski, Ohad Fried, and Xi Yin. Spatext: Spatio-textual representation for con- trollable image generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023. 5

    work page 2023

  4. [4]

    Kinstyle: A strong baseline photorealistic kinship face syn- thesis with an optimized stylegan encoder

    Li-Chen Cheng, Shu-Chuan Hsu, Pin-Hua Lee, Hsiu-Chieh Lee, Che-Hsien Lin, Jun-Cheng Chen, and Chih-Yu Wang. Kinstyle: A strong baseline photorealistic kinship face syn- thesis with an optimized stylegan encoder. InProceedings of the Asian Conference on Computer Vision (ACCV), 2022. 2, 3, 4, 5, 6, 8, 1

    work page 2022

  5. [5]

    Human facial shape and size heritability and genetic correlations

    Joanne B Cole, Mange Manyama, Jacinda R Larson, Denise K Liberton, Tracey M Ferrara, Sheri L Riccardi, Mao Li, Washington Mio, Ophir D Klein, Stephanie A Santorico, et al. Human facial shape and size heritability and genetic correlations. Genetics, 205(2):967–978, 2017. 2

    work page 2017

  6. [6]

    Gantastic: Gan-based transfer of interpretable directions for disentan- gled image editing in text-to-image diffusion models

    Yusuf Dalva, Hidir Yesiltepe, and Pinar Yanardag. Gantastic: Gan-based transfer of interpretable directions for disentan- gled image editing in text-to-image diffusion models. arXiv preprint arXiv:2403.19645, 2024. 2

    arXiv 2024

  7. [7]

    Social perception of facial resemblance in humans

    Lisa M DeBruine, Benedict C Jones, Anthony C Little, and David I Perrett. Social perception of facial resemblance in humans. Archives of sexual behavior, 37:64–77, 2008. 5, 7

    work page 2008

  8. [8]

    Arcface: Additive angular margin loss for deep face recognition

    Jiankang Deng, Jia Guo, Niannan Xue, and Stefanos Zafeiriou. Arcface: Additive angular margin loss for deep face recognition. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR) ,

    work page

Show all 41 references
  1. [9]

    Hyperdiffusion: Generating implicit neural fields with weight-space diffusion

    Ziya Erkoc ¸, Fangchang Ma, Qi Shan, Matthias Nießner, and Angela Dai. Hyperdiffusion: Generating implicit neural fields with weight-space diffusion. In Proceedings of the In- ternational Conference on Computer Vision (ICCV) , 2023. 3

    2023

  2. [10]

    Concept sliders: Lora adap- tors for precise control in diffusion models

    Rohit Gandikota, Joanna Materzy ´nska, Tingrui Zhou, Anto- nio Torralba, and David Bau. Concept sliders: Lora adap- tors for precise control in diffusion models. arXiv preprint arXiv:2311.12092, 2023. 2

    2023 arXiv

  3. [11]

    What will your child look like? dna-net: Age and gender aware kin face synthesizer

    Pengyu Gao, Siyu Xia, Joseph Robinson, Junkang Zhang, Chao Xia, Ming Shao, and Yun Fu. What will your child look like? dna-net: Age and gender aware kin face synthesizer. Proceedings of the International Conference on Multimedia and Expo (ICME), 2021. 2, 3

    2021

  4. [12]

    Generative adversarial networks

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial networks. Commu- nications of the ACM, 2020. 2

    2020

  5. [13]

    Ms-celeb-1m: A dataset and benchmark for large-scale face recognition

    Yandong Guo, Lei Zhang, Yuxiao Hu, Xiaodong He, and Jianfeng Gao. Ms-celeb-1m: A dataset and benchmark for large-scale face recognition. In Proceedings of the European Conference on Computer Vision (ECCV), 2016. 7

    2016

  6. [14]

    Classifier-free diffusion guidance

    Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598, 2022. 2, 5

    2022 arXiv

  7. [15]

    Denoising diffu- sion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffu- sion probabilistic models. Advances in Neural Information Processing Systems, 2020. 2, 4

    2020

  8. [16]

    Stylefusion: A generative model for disentangling spatial segments

    Omer Kafri, Or Patashnik, Yuval Alaluf, and Daniel Cohen- Or. Stylefusion: A generative model for disentangling spatial segments. arXiv preprint arXiv:2107.07437, 2021. 2

    2021 arXiv

  9. [17]

    A style-based generator architecture for generative adversarial networks

    Tero Karras, Samuli Laine, and Timo Aila. A style-based generator architecture for generative adversarial networks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019. 2, 1

    2019

  10. [18]

    Analyzing and improv- ing the image quality of stylegan

    Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. Analyzing and improv- ing the image quality of stylegan. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020. 2, 4, 1

    2020

  11. [19]

    Alias-free generative adversarial networks

    Tero Karras, Miika Aittala, Samuli Laine, Erik H ¨ark¨onen, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. Alias-free generative adversarial networks. Advances in Neural Infor- mation Processing Systems, 34:852–863, 2021. 2

    2021

  12. [20]

    Style- gene: Crossover and mutation of region-level facial genes for kinship face synthesis

    Hao Li, Xianxu Hou, Zepeng Huang, and Linlin Shen. Style- gene: Crossover and mutation of region-level facial genes for kinship face synthesis. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023. 2, 3, 5, 6, 8

    2023

  13. [21]

    When stylegan meets stable diffusion: a w+ adapter for person- alized image generation

    Xiaoming Li, Xinyu Hou, and Chen Change Loy. When stylegan meets stable diffusion: a w+ adapter for person- alized image generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 2187–2196, 2024. 2, 4

    2024

  14. [22]

    Styledna: A high-fidelity age and gender aware kinship face synthe- sizer

    Che-Hsien Lin, Hung-Chun Chen, Li-Chen Cheng, Shu- Chuan Hsu, Jun-Cheng Chen, and Chih-Yu Wang. Styledna: A high-fidelity age and gender aware kinship face synthe- sizer. In Proceedings of the IEEE International Conference on Automatic Face and Gesture Recognition (FG), pages ...

    2021

  15. [23]

    Kartik Narayan, Vibashan VS, Rama Chellappa, and Vishal M. Patel. Facexformer: A unified transformer for fa- cial analysis. arXiv preprint arXiv:2403.12960, 2024. 1

    2024 arXiv

  16. [24]

    Kinshipgan: Synthesizing of kinship faces from family photos by regularizing a deep face network

    Savas Ozkan and Akin Ozkan. Kinshipgan: Synthesizing of kinship faces from family photos by regularizing a deep face network. In Proceedings of the IEEE international confer- ence on image processing (ICIP), 2018. 2, 3

    2018

  17. [25]

    Venkatesh Babu

    Rishubh Parihar, Sachidanand VS, Sabariswaran Mani, Te- jan Karmali, and R. Venkatesh Babu. Precisecontrol : En- hancing text-to-image diffusion models with fine-grained at- tribute control. In Proceedings of the Eurpean Conference on Computer Vision (ECCV), 2024. 2, 4

    2024

  18. [26]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR) ,

  19. [27]

    Kfc: Kinship verification with fair contrastive loss and multi-task learning, 2023

    Jia Luo Peng, Keng Wei Chang, and Shang-Hong Lai. Kfc: Kinship verification with fair contrastive loss and multi-task learning, 2023. 2

    2023

  20. [28]

    clip2latent

    Justin NM Pinkney and Chuan Li. clip2latent. arXiv preprint arXiv:2210.02347, 2022. 3

    2022 arXiv

  21. [29]

    Tri-subject kinship verification: Understanding the core of a family

    Xiaoqian Qin, Xiaoyang Tan, and Songcan Chen. Tri-subject kinship verification: Understanding the core of a family. IEEE Transactions on Multimedia, 2015. 5, 6

    2015

  22. [30]

    Robinson, Ming Shao, Yue Wu, and Yun Fu

    Joseph P. Robinson, Ming Shao, Yue Wu, and Yun Fu. Fam- ilies in the wild (fiw): Large-scale kinship image database and benchmarks. In Proceedings of the ACM on Multimedia Conference (MM), 2016. 5, 6

    2016

  23. [31]

    Interfacegan: Interpreting the disentangled face representa- tion learned by gans

    Yujun Shen, Ceyuan Yang, Xiaoou Tang, and Bolei Zhou. Interfacegan: Interpreting the disentangled face representa- tion learned by gans. IEEE transactions on pattern analysis and machine intelligence, 2020. 2

    2020

  24. [32]

    Familygan: Generating kin face images using generative adversarial net- works

    Raunak Sinha, Mayank Vatsa, and Richa Singh. Familygan: Generating kin face images using generative adversarial net- works. In Proceedings of the European Conference on Com- puter Vision Workshop (ECCVW), pages 297–311, 2020. 2, 3

    2020

  25. [33]

    Deep unsupervised learning using nonequilibrium thermodynamics

    Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. In Proceedings of the In- ternational Conference on Machine Learning (ICML), 2015. 4

    2015

  26. [34]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502, 2020. 2, 4

    2010 arXiv

  27. [35]

    Designing an encoder for stylegan image manipulation

    Omer Tov, Yuval Alaluf, Yotam Nitzan, Or Patashnik, and Daniel Cohen-Or. Designing an encoder for stylegan image manipulation. ACM Transactions on Graphics (TOG), 40(4): 1–14, 2021. 1

    2021

  28. [36]

    Stylespace analysis: Disentangled controls for stylegan image genera- tion

    Zongze Wu, Dani Lischinski, and Eli Shechtman. Stylespace analysis: Disentangled controls for stylegan image genera- tion. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), 2021. 2, 4

    2021

  29. [37]

    A latent transformer for disentangled face editing in images and videos

    Xu Yao, Alasdair Newson, Yann Gousseau, and Pierre Hel- lier. A latent transformer for disentangled face editing in images and videos. In Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision (ICCV) , pages 13789–13798, 2021. 2

    2021

  30. [38]

    Deep fusion siamese network for automatic kinship verification

    Jun Yu, Mengyan Li, Xinlong Hao, and Guochen Xie. Deep fusion siamese network for automatic kinship verification. In Proceedings of the IEEE International Conference on Auto- matic Face and Gesture Recognition (FG) , pages 892–899,

  31. [39]

    Styleavatar3d: Leveraging image-text diffusion models for high-fidelity 3d avatar generation

    Chi Zhang, Yiwen Chen, Yijun Fu, Zhenglin Zhou, Gang Yu, Billzb Wang, Bin Fu, Tao Chen, Guosheng Lin, and Chun- hua Shen. Styleavatar3d: Leveraging image-text diffusion models for high-fidelity 3d avatar generation. arXiv preprint arXiv:2305.19012, 2023. 3

    2023 arXiv

  32. [40]

    Childpredictor: A child face pre- diction framework with disentangled learning

    Yuzhi Zhao, Lai-Man Po, Xuehui Wang, Qiong Yan, Wei Shen, Yujia Zhang, Wei Liu, Chun-Kit Wong, Chiu-Sing Pang, Weifeng Ou, et al. Childpredictor: A child face pre- diction framework with disentangled learning. IEEE Trans- actions on Multimedia, 2022. 2, 3, 5, 6, 8

    2022

  33. [41]

    Towards robust blind face restora- tion with codebook lookup transformer

    Shangchen Zhou, Kelvin Chan, Chongyi Li, and Chen Change Loy. Towards robust blind face restora- tion with codebook lookup transformer. Advances in Neural Information Processing Systems, 35:30599–30611, 2022. 5, 8, 4 10 StyleDiT: A Unified Framework for Diverse Child and Partn...

    2022

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.