Pith. sign in

REVIEW 4 major objections 6 minor 72 references

MyTimeMachine: Personalized Facial Age Transformation

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read A small personal photo album—as few as 50 images spanning decades—is enough to synthesize how a specific face ages, both into the past and the future.

desk verdict Novel adapter-based personalization for face aging, but the headline identity metric is the training objective, so the central quantitative claim doesn't hold as reported. read the letter →

arxiv 2411.14521 v2 pith:CL5J2T3J submitted 2024-11-21 cs.CV

classification cs.CV
keywords facialagetransformationpersonalizationprogressionregressionadapternetworkStyleGAN2latentspaceeditingidentitypreservation
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 argues that accurate facial aging for a specific person does not require a giant dataset—only a few dozen photos of that person taken over 20 to 40 years. The method, MyTimeMachine (MyTM), starts from a global aging model that knows how faces age on average and adds a lightweight person-specific adapter network that corrects that average trajectory using the personal photo collection. With as few as 50 images, the resulting re-aged faces both land at the requested target age and resemble how that individual actually looked at that age, and the same personalization extends to video. This matters because movie de-aging, missing-person aging, and personal time-machine apps could then work from an ordinary photo album instead of generic templates that produce plausible but wrong faces.

What carries the argument

The load-bearing object is the Age Adapter Network, a family of 18 style-wise MLPs (plus a shared global MLP and an aging MLP) that reads SAM's W+ latent code and the target age and emits a per-style latent offset ΔW+. The offset is added to the global code before the frozen StyleGAN2 decoder, so personalization never retrains the generator or the age encoder. The argument is carried equally by the three losses: the personalized aging loss uses an ArcFace reference set near the target age to anchor identity without relying on a possibly biased global age classifier; the extrapolation regularization replays SAM's output for out-of-range target ages to prevent overfitting to the training age boundaries; and the adaptive w-norm regularization, with a cosine schedule on the age gap, keeps latent codes near the data manifold for editing while staying close enough to SAM's inversion for fidelity.

What would settle it

Hold out a person's childhood photos, train MyTM only on photos from ages 50–70, de-age a 70-year-old test photo to age 10, and compare the output side-by-side with the withheld childhood photo; if a substantial fraction of subjects produce outputs that do not resemble the withheld early photos—measured by ArcFace similarity against the actual childhood reference set falling below the paper's reported interpolation levels—then the extrapolation claim is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a personalized facial age transformation can be learned from a small longitudinal photo collection (ideally 50 images spanning 20–40 years) by combining two sources of knowledge: a global aging prior learned from many identities, and a personal adapter that corrects that prior for one individual. MyTM builds on SAM, a StyleGAN2-based encoder that maps an input face and a target age into the W+ latent space. A new Age Adapter Network—a set of MLPs—takes SAM's predicted latent code and the target age and outputs a latent offset; adding that offset to the global code and decoding with the frozen StyleGAN2 generator produces the personalized re-aged face. Three losses drive personalization: a personalized aging loss that pulls the output's identity features toward the person's own photos near the target age (using maximum ArcFace cosine similarity over a reference set), an extrapolation regularization that keeps outputs close to the global model when the target age lies outside the training range (experience replay), and an adaptive w-norm regularization that schedules proximity to the mean latent code as a cosine function of the input-to-target age gap to balance inversion accuracy and editability. On a curated 12-celebrity longitudinal dataset, MyTM reports age errors of 7.7 (regression) and 6.3 (progression) MAE with identity similarities around 0.76–0.78, outperforming SAM, CUSP, AgeTransGAN, FADING, and naive personalization baselines such as DreamBooth fine-tuning.

Load-bearing premise

The personal photo collection must genuinely represent how the person aged: ages come from metadata and old photos are restored with an external enhancer (GFPGAN), so if the collection is retouched, professionally lit, or biased in pose or lighting, the adapter may learn artifacts rather than the true aging trajectory, and the claimed resemblance at the target age fails.

Editorial extensions

If this is right

  • With roughly 50 personal photos spanning decades, one can generate de-aged actors for VFX without generic aging templates, and age people forward for missing-person or forensic use.
  • The same trained model extrapolates beyond the photo collection's age range (e.g., trained on ages 20–40, aged to 100) with a reported age MAE of 6.3, so the personalization is not limited to interpolation.
  • Personalization transfers to video by re-aging a single keyframe and face-swapping into the rest of the clip, giving temporally consistent results in under 5 minutes versus over 3 hours for direct video editing baselines.
  • Naive personalization of global models (DreamBooth-style fine-tuning) overfits to the training age range, whereas MyTM's combination of personal and global priors avoids that failure.
  • The improvement holds for non-celebrity YouTube subjects, not just curated celebrity photos, indicating the approach is not limited to professionally captured faces.

Reading between the lines

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

  • The adapter-plus-reference-set pattern suggests a general recipe for personalizing other attribute editors (e.g., weight, hairstyle, or skin health) with a handful of exemplars; the paper demonstrates it for age, but the machinery is attribute-agnostic.
  • The adaptive w-norm cosine schedule implies that a continuous trade-off curve between inversion fidelity and editability can be navigated per-example; a testable extension would be to replace the cosine with a learned schedule or apply the same schedule to other latent-editing tasks.
  • The video result suggests temporal consistency can be decoupled from the aging model itself by delegating it to a face-swapper, and if face-swapping improves, the same pipeline could handle other identity-preserving edits such as expression or lighting changes.
  • A practical caution follows from the weakest assumption: the method's reliability on real photo albums depends on restoration not erasing wrinkles or sagging; a robustness test would re-run the experiments with different restoration settings to quantify this sensitivity.
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 / 6 minor

Summary. The paper proposes MyTimeMachine (MyTM), a personalized facial age transformation method built on the global aging encoder SAM and a fixed StyleGAN2 decoder. For a given person, MyTM trains a per-identity adapter network that predicts an offset to SAM's W+ latent code conditioned on the target age, using three losses: a personalized aging loss that maximizes ArcFace similarity to reference images of the same person near the target age (Eq. 5), an extrapolation regularization that keeps outputs close to SAM's global prediction outside the training age range (Eq. 6), and an adaptive w-norm regularizer (Eq. 7). The method is evaluated for age regression (de-aging) and progression on a curated set of 12 celebrities and 5 non-celebrities, with an additional video re-aging extension based on face-swapping. The paper claims that with 10-50 personal images spanning 20-40 years, MyTM produces re-aged images that are both age-accurate and identity-preserving, outperforming global and naive-personalization baselines.

Significance. Practical significance is potentially high: it addresses an applied problem (VFX, forensic aging) where personal photo collections are available, and the proposed adapter-plus-regularization recipe is simple and clearly specified. The ablation in Fig. 8 and the dataset-size study in Fig. 7 are useful and support the claim that each component contributes and that around 50 images is a reasonable operating point. The video extension via face-swapping is pragmatic, and the user-study preference data, though small, is directionally favorable. However, the identity-preservation evidence is currently undermined by the fact that the headline metric ID_sim (Eq. 8) is the same quantity that the personalized aging loss (Eq. 5) optimizes, and the quantitative evaluation rests on aggregate means over 8-12 subjects without error bars. These issues must be resolved before the central claims can be considered established.

major comments (4)
  1. [§3.3.1, Eq. (5); §4.1.4, Eq. (8)] The personalized aging loss is L_pers-age = 1 - max_j <R(y^p_tgt), R(x_j)> over a reference set D_tgt defined by a_tgt +/- 3 years, and the reported identity metric is ID_sim = max_j <R(y_tgt), R(x_j)> over the same kind of reference window, averaged over target ages. The only stated distinction is that the evaluation reference images are not used in training, but they come from the same curated collection and the same GFPGAN restoration/alignment pipeline (Appendix B). Consequently, MyTM is directly trained to optimize the very quantity on which its identity-preservation advantage is reported (e.g., Table 1: 0.67 vs 0.60; Table 2: 0.70 vs 0.62), while the baselines are not trained with this objective. This is load-bearing because identity preservation is one of the two headline claims. Please re-evaluate with an identity metric that is not part of the training objective (for example, a different face-recognition backbone or human 'same identity' judgments), and report per-celebrity results with variance. The max over the reference set can also be inflated by a single outlier reference, so a mean or median over the reference set should be reported alongside it.
  2. [§4.1.1, Appendix B] The quantitative claims are supported by aggregate means over 12 celebrities (8-10 per task) with no standard deviations, confidence intervals, significance tests, or per-subject breakdowns, and the dataset is not released. The ablation in Fig. 8 is additionally performed on a single celebrity (Al Pacino). The preprocessing pipeline restores low-quality and grayscale images with GFPGAN, which can alter aging texture (wrinkles, sagging), and ages are inferred from photo metadata where available; both steps can bias the personalized aging trajectory. Please report per-celebrity numbers with error bars, clarify how age labels were obtained and verified, and make the dataset or a reproducible subset available. Without these, the general claim that 'as few as 50 images' suffices is not established beyond anecdote.
  3. [Appendix E, Fig. 11] The paper's own comparison shows that SAM Pers. f.t. + MyStyle achieves a high ID_sim (0.60) while producing images described as 'adversarial examples for arcface' with visibly poor quality. This is internal evidence that ID_sim can be inflated by non-identity factors, so ID_sim alone cannot carry the identity-preservation claim. The user study in Fig. 4 asks participants to match the reference images while preserving style, which conflates target-age resemblance with identity preservation, and it reports no significance tests. Please add a human evaluation that isolates identity (e.g., forced-choice 'same person?' at matched apparent age) and report the number of participants, per-condition counts, and confidence intervals.
  4. [§4.2.1, Tables 1-2] The reported aggregates mix interpolation and extrapolation targets. For age regression, Ours (30~70) evaluated at a_tgt <= 70 includes target ages within the training range (30-70) and extrapolation ages (0-20); for age progression, targets 40-100 include the boundary age 40 and extrapolation beyond it. Since extrapolation is a central claimed advantage, please report interpolation and extrapolation performance separately for both tasks, in addition to the pooled numbers.
minor comments (6)
  1. [§3.3] The section title 'Loss Funtions' should be 'Loss Functions'.
  2. [Fig. 10 caption] The caption says 'The input test images match those in Fig. 10 (top)' but should refer to Fig. 9 for the pre-trained baseline comparison.
  3. [Eq. (7)] The adaptive w-norm schedule is written as lambda_reg(Delta_age) = 1 - <pi * Delta_age / 100>; please specify whether <·> denotes cosine, state the intended domain of Delta_age, and confirm the sign and range of lambda_reg, since the text describes it as a cosine function.
  4. [Abstract and Section 1] The abstract says 'as few as 50 images' while the introduction says 'as few as 10, ideally 50'; please align these statements with the dataset-size results in Fig. 7.
  5. [Table 5] Age ranges such as '01~76' and '03~96' should be written without leading zeros (e.g., '1~76', '3~96').
  6. [Appendix B] For many celebrity photos, capture-time metadata is unavailable; state how age labels were obtained or verified beyond metadata.

Circularity Check

1 steps flagged · score 6.0 of 10

The reported identity metric ID_sim is the same max-ArcFace similarity that the personalized aging loss optimizes (Eq. 5 vs. Eq. 8), so the headline identity-preservation advantage is partly built into the training objective by construction.

  1. fitted input called prediction [Section 3.3.1 Eq. (5); Section 4.1.4 Eq. (8)]
    "Lpers-age = 1−𝑚𝑎𝑥{⟨𝑅(𝑦𝑝 tgt),𝑅(𝑥 𝑗)⟩}𝑀 𝑗=1 (5) ... ID𝑠𝑖𝑚(𝑦tgt)=𝑚𝑎𝑥{⟨𝑅(𝑦 tgt),𝑅(𝑥 𝑗)⟩}𝑀 𝑗=1 (8) where 𝑅(·) is a pretrained arcface [Deng et al. 2022] network for facial feature recognition and 𝑥𝑗 belongs in reference image set near the target age (𝑎tgt± 3-years)."

    The personalized aging loss in Eq. (5) minimizes 1 minus the maximum ArcFace cosine similarity between the re-aged image and reference images within ±3 years of the target age. The identity-preservation metric ID_sim in Eq. (8) is exactly that same maximum ArcFace cosine similarity against reference images within ±3 years of the target age. Thus the ID_sim values reported in Tables 1 and 2 and the claimed 11.7% improvement over global baselines are measurements of the training objective itself; baselines such as SAM and FADING are not trained with this objective, so the comparison is biased toward MyTM by construction. Appendix E reinforces this concern: SAM Pers. ft.

full rationale

The paper's architectural contribution—a personalized adapter trained on a small personal photo collection with extrapolation regularization and adaptive w-norm regularization—is not itself circular. The age-accuracy metric Age_MAE uses the external FP-Age detector, the user studies are human preferences, and the extrapolation and w-norm losses are defined relative to SAM's output and the W-space norm, all independent of the headline identity metric. However, the identity-preservation claim as quantified is partially circular: Eq. (5) defines the personalized aging loss as 1 minus the maximum ArcFace cosine similarity to target-age reference images, and Eq. (8) defines ID_sim as exactly that maximum similarity to the same type of reference set. The reported identity advantage therefore measures how well the model optimized its own training objective, not an independent test of re-aging fidelity. Appendix E's admission that high ID_sim can correspond to adversarial examples for ArcFace further weakens the metric's independence. Since the age-accuracy and user-study components remain externally grounded, the circularity is partial rather than total; the score reflects that one of the paper's central quantitative claims reduces, in large part, to the training loss.

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

The method introduces no new physical entities. Its load-bearing components are hand-chosen hyperparameters (all loss weights set to 1, the cosine w-norm schedule, the +/-3-year reference window, and the 50-image dataset size) plus domain assumptions inherited from SAM, StyleGAN2, arcface, FP-Age, and the curated celebrity photo collection. The most consequential coupling is that the training objective and the identity evaluation metric coincide, which inflates the apparent identity gain.

free parameters (4)
  • Adaptive w-norm cosine schedule = 1 - cos(pi * Delta_age / 100)
    The regularization weight in Eq. 7 is an ad hoc cosine function of age difference. The 100 scaling and the cosine shape are hand-chosen, not derived, and directly control the inversion-editability trade-off.
  • Reference age window = +/- 3 years
    The reference set D_tgt is defined as images within +/-3 years of the target age (Sec. 3.3.1). This window size is hand-picked and affects both the training loss and the ID_sim evaluation metric.
  • Loss weights = lambda_pers-age = 1, lambda_reg-extra = 1, lambda_reg = 1
    All three introduced losses use weight 1 (Appendix C), inherited from SAM's original weighting. No sensitivity analysis is reported.
  • Training set size = 50 images, with 10 and 100 in ablation
    Fifty images per person is the default. The ablation in Fig. 7 shows clear improvement from 10 to 50 and saturation at 100, making the choice defensible but still a hand choice.
assumptions (4)
  • domain assumption SAM provides an accurate, well-calibrated global aging prior in StyleGAN2 W+ space
    The entire method builds on SAM (Alaluf et al. 2021) as a frozen backbone (Sec. 3.1). If SAM's global aging directions are biased or poorly editable, the adapter operates on a bad base.
  • domain assumption ArcFace cosine similarity to reference photos measures resemblance at the target age
    Used both as the training loss (Eq. 5) and the evaluation metric (Eq. 8). The paper itself notes arcface has age dependence (Sec. 4.1.4), so this is a partial proxy for true visual resemblance.
  • domain assumption FP-Age predicts age accurately enough to serve as ground truth for Age_MAE
    All age accuracy numbers are |FP-Age(output) - a_tgt|. If FP-Age is biased for the tested celebrities or age ranges, Age_MAE is systematically wrong.
  • domain assumption Longitudinal celebrity photos with metadata ages reflect genuine aging, and GFPGAN restoration of old images does not distort aging cues
    Dataset construction (Sec. 4.1.1, Appendix B) relies on public images with estimated capture dates and restores low-quality old images. Restored or retouched photos can erase genuine aging texture and inject a modern appearance, biasing the learned personal trajectory.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MyTimeMachine: Personalized Facial Age Transformation." pith.science (2026). https://pith.science/paper/CL5J2T3J

@misc{pith2026241114521,
  author       = {Pith},
  title        = {Pith review of: MyTimeMachine: Personalized Facial Age Transformation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CL5J2T3J}},
  note         = {Machine review of arXiv:2411.14521}
}
abstract

Facial aging is a complex process, highly dependent on multiple factors like gender, ethnicity, lifestyle, etc., making it extremely challenging to learn a global aging prior to predict aging for any individual accurately. Existing techniques often produce realistic and plausible aging results, but the re-aged images often do not resemble the person's appearance at the target age and thus need personalization. In many practical applications of virtual aging, e.g. VFX in movies and TV shows, access to a personal photo collection of the user depicting aging in a small time interval (20$\sim$40 years) is often available. However, naive attempts to personalize global aging techniques on personal photo collections often fail. Thus, we propose MyTimeMachine (MyTM), which combines a global aging prior with a personal photo collection (using as few as 50 images) to learn a personalized age transformation. We introduce a novel Adapter Network that combines personalized aging features with global aging features and generates a re-aged image with StyleGAN2. We also introduce three loss functions to personalize the Adapter Network with personalized aging loss, extrapolation regularization, and adaptive w-norm regularization. Our approach can also be extended to videos, achieving high-quality, identity-preserving, and temporally consistent aging effects that resemble actual appearances at target ages, demonstrating its superiority over state-of-the-art approaches.

Figures

Figures reproduced from arXiv: 2411.14521 by the authors.

Figure 1
Figure 1. We introduce MyTimeMachine (MyTM) to perform personalized age regression (top) and progression (bottom) by training a person-specific aging [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Given an input face of Oprah Winfrey at 70 years old, our adapter re-ages her face to resemble her appearance at 30, while preserving the style of the input image. To achieve personalized re-aging, we collect ∼50 images of an individual across different ages and train an adapter network that updates the latent code generated by the global age encoder SAM. Our adapter preserves identity during interpolation when the … view at source ↗
Figure 3
Figure 3. Performance of age transformation techniques for age regression (first two rows) and age progression (last two rows). The first column shows the input [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (11 more)
Figure 9
Figure 9. Figure 9: Additionally, visual results are compared against personalized [PITH_FULL_IMAGE:figures/full_fig_p006_9.png]
Figure 5
Figure 5. Figure 5: Naive pasting of MyTM’s re-aged face onto a video frame of [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Face-swapping for video re-aging on a clip of [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 4
Figure 4. Figure 4: User study comparing our method with baselines—FADING, Age [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 7
Figure 7. Figure 7: Effect of training dataset size D on personalization. MyTM is trained on ages 30∼70 and tested for 𝑎tgt ≤ 70. Visual examples of Robert De Niro are shown at the top, with quantitative results displayed below. MyTM achieves personalized re-aging with as few as 10 images…
Figure 8
Figure 8. Figure 8: Contributions of our proposed loss functions and the adapter network [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: Performance of age transformation techniques for age regression (top) and age progression (bottom). For age regression, MyTM (Ours) is trained across [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: Performance of personalized age transformation techniques for age regression (top) and age progression (bottom). The input test images match [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]
Figure 11
Figure 11. Figure 11: We compare MyTM (Ours) with naive personalization techniques: [PITH_FULL_IMAGE:figures/full_fig_p015_11.png]
Figure 12
Figure 12. Figure 12: Limitations of MyTM. Our method may struggle with accessories (e.g., glasses), as these elements are not consistently handled by the e4e encoder [Tov [PITH_FULL_IMAGE:figures/full_fig_p016_12.png]
Figure 13
Figure 13. Figure 13: Visual results of FADING using identical input and inference code. The instability in age transformation arises from the optimization of NTI [Mokady [PITH_FULL_IMAGE:figures/full_fig_p016_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

72 extracted references · 43 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    Adaptive IMLE for Few -shot Pretraining -free Generative Modelling

    Mehran Aghabozorgi, Shichong Peng, and Ke Li. Adaptive IMLE for Few -shot Pretraining -free Generative Modelling . 2023

  3. [3]

    Only a matter of style: age transformation using a style-based regression model

    Yuval Alaluf, Or Patashnik, and Daniel Cohen-Or. Only a matter of style: age transformation using a style-based regression model. ACM Transactions on Graphics, 40 0 (4): 0 1--12, 2021. TLDR: An image-to-image translation method that learns to directly encode real facial images into the latent space of a pre-trained unconditional GAN subject to a given agi...

  4. [4]

    Semantic Photo Manipulation with a Generative Image Prior

    David Bau, Hendrik Strobelt, William Peebles, Jonas Wulff, Bolei Zhou, Jun-Yan Zhu, and Antonio Torralba. Semantic Photo Manipulation with a Generative Image Prior . ACM Transactions on Graphics, 38 0 (4): 0 1--11, 2019. arXiv:2005.07727 [cs] TLDR: This paper adapts the image prior learned by GANs to image statistics of an individual image and can accurat...

  5. [5]

    Continuous, Subject - Specific Attribute Control in T2I Models by Identifying Semantic Directions , 2024

    Stefan Andreas Baumann, Felix Krause, Michael Neumayr, Nick Stracke, Vincent Tao Hu, and Björn Ommer. Continuous, Subject - Specific Attribute Control in T2I Models by Identifying Semantic Directions , 2024. arXiv:2403.17064 [cs] TLDR: It is shown that there exist directions in the commonly used token-level CLIP text embeddings that enable fine-grained su...

  6. [6]

    Anand Bhattad, Viraj Shah, Derek Hoiem, and D. A. Forsyth. Make It So : Steering StyleGAN for Any Image Inversion and Editing , 2023. arXiv:2304.14403 [cs]

  7. [7]

    SimSwap : An Efficient Framework For High Fidelity Face Swapping , 2021

    Renwang Chen, Xuanhong Chen, Bingbing Ni, and Yanhao Ge. SimSwap : An Efficient Framework For High Fidelity Face Swapping , 2021. arXiv:2106.06340

  8. [8]

    Face Aging via Diffusion -based Editing , 2023

    Xiangyi Chen and Stéphane Lathuilière. Face Aging via Diffusion -based Editing , 2023. arXiv:2309.11321 [cs]

Show all 72 references
  1. [9]

    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 Conference on Computer Vision and Pattern Recognition, pages 4690--4699, 2019

  2. [10]

    ArcFace : Additive Angular Margin Loss for Deep Face Recognition

    Jiankang Deng, Jia Guo, Jing Yang, Niannan Xue, Irene Kotsia, and Stefanos Zafeiriou. ArcFace : Additive Angular Margin Loss for Deep Face Recognition . IEEE Transactions on Pattern Analysis and Machine Intelligence, 44 0 (10): 0 5962--5979, 2022. arXiv:1801.07698 [cs]

  3. [11]

    AgingMapGAN ( AMGAN ): High - Resolution Controllable Face Aging with Spatially - Aware Conditional GANs , 2020

    Julien Despois, Frederic Flament, and Matthieu Perrot. AgingMapGAN ( AMGAN ): High - Resolution Controllable Face Aging with Spatially - Aware Conditional GANs , 2020. arXiv:2008.10960

  4. [12]

    Efros, and Kfir Aberman

    Amil Dravid, Yossi Gandelsman, Kuan-Chieh Wang, Rameen Abdal, Gordon Wetzstein, Alexei A. Efros, and Kfir Aberman. Interpreting the Weight Space of Customized Diffusion Models , 2024. arXiv:2406.09413 [cs] TLDR: It is shown that inverting a single image into this space reconst...

  5. [13]

    Yun Fu, Guodong Guo, and Thomas S. Huang. Age synthesis and estimation via faces: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 32 0 (11): 0 1955--1976, 2010

  6. [14]

    Information Bottleneck Disentanglement for Identity Swapping

    Gege Gao, Huaibo Huang, Chaoyou Fu, Zhaoyang Li, and Ran He. Information Bottleneck Disentanglement for Identity Swapping . pages 3404--3413, 2021

  7. [15]

    Custom Structure Preservation in Face Aging , 2022

    Guillermo Gomez-Trenado, Stéphane Lathuilière, Pablo Mesejo, and Óscar Cordón. Custom Structure Preservation in Face Aging , 2022. arXiv:2207.11025 [cs]

  8. [16]

    S2GAN : Share Aging Factors Across Ages and Share Aging Trends Among Individuals

    Zhenliang He, Meina Kan, Shiguang Shan, and Xilin Chen. S2GAN : Share Aging Factors Across Ages and Share Aging Trends Among Individuals . In 2019 IEEE / CVF International Conference on Computer Vision ( ICCV ) , pages 9439--9448, Seoul, Korea (South), 2019. IEEE. TLDR: The pr...

  9. [17]

    Prompt-to- Prompt Image Editing with Cross Attention Control , 2022

    Amir Hertz, Ron Mokady, Jay Tenenbaum, Kfir Aberman, Yael Pritch, and Daniel Cohen-Or. Prompt-to- Prompt Image Editing with Cross Attention Control , 2022. arXiv:2208.01626 TLDR: This paper analyzes a text-conditioned model in depth and observes that the cross-attention layers...

  10. [18]

    AgeTransGAN for Facial Age Transformation with Rectified Performance Metrics

    Gee-Sern Hsu, Rui-Cang Xie, Zhi-Ting Chen, and Yu-Hong Lin. AgeTransGAN for Facial Age Transformation with Rectified Performance Metrics . In Computer Vision – ECCV 2022 , pages 580--595. Springer Nature Switzerland, Cham, 2022. Series Title: Lecture Notes in Computer Science

  11. [19]

    GANSpace : Discovering Interpretable GAN Controls , 2020

    Erik Härkönen, Aaron Hertzmann, Jaakko Lehtinen, and Sylvain Paris. GANSpace : Discovering Interpretable GAN Controls , 2020. arXiv:2004.02546 [cs]

  12. [20]

    iperov/ DeepFaceLive , 2024

    iperov. iperov/ DeepFaceLive , 2024. original-date: 2020-12-15T12:19:22Z

  13. [21]

    A Style - Based Generator Architecture for Generative Adversarial Networks , 2019

    Tero Karras, Samuli Laine, and Timo Aila. A Style - Based Generator Architecture for Generative Adversarial Networks , 2019. arXiv:1812.04948 [cs, stat]

  14. [22]

    Analyzing and Improving the Image Quality of StyleGAN , 2020 a

    Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. Analyzing and Improving the Image Quality of StyleGAN , 2020 a . arXiv:1912.04958 [cs, eess, stat]

  15. [23]

    Analyzing and improving the image quality of StyleGAN

    Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. Analyzing and improving the image quality of StyleGAN . In Proc. CVPR, 2020 b

  16. [24]

    Ira Kemelmacher-Shlizerman, Supasorn Suwajanakorn, and Steven M. Seitz. Illumination- Aware Age Progression . pages 3334--3341, 2014

  17. [25]

    Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, Demis Hassabis, Claudia Clopath, Dharshan Kumaran, and Raia Hadsell

    James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A. Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, Demis Hassabis, Claudia Clopath, Dharshan Kumaran, and Raia Hadsell. Overcoming catastrophic forgetting i...

  18. [26]

    Diffusion Models already have a Semantic Latent Space , 2023

    Mingi Kwon, Jaeseok Jeong, and Youngjung Uh. Diffusion Models already have a Semantic Latent Space , 2023. arXiv:2210.10960 [cs] TLDR: Asyrp is proposed which discovers the semantic latent space in frozen pretrained diffusion models, named h-space, which has nice properties fo...

  19. [27]

    Countering Language Drift via Visual Grounding , 2019

    Jason Lee, Kyunghyun Cho, and Douwe Kiela. Countering Language Drift via Visual Grounding , 2019. arXiv:1909.04499 [cs]

  20. [28]

    Pluralistic Aging Diffusion Autoencoder , 2023

    Peipei Li, Rui Wang, Huaibo Huang, Ran He, and Zhaofeng He. Pluralistic Aging Diffusion Autoencoder , 2023. arXiv:2303.11086 [cs]

  21. [29]

    Continuous Face Aging via Self -estimated Residual Age Embedding

    Zeqi Li, Ruowei Jiang, and Parham Aarabi. Continuous Face Aging via Self -estimated Residual Age Embedding . In 2021 IEEE / CVF Conference on Computer Vision and Pattern Recognition ( CVPR ) , pages 15003--15012, Nashville, TN, USA, 2021. IEEE

  22. [30]

    FP - Age : Leveraging Face Parsing Attention for Facial Age Estimation in the Wild , 2022

    Yiming Lin, Jie Shen, Yujiang Wang, and Maja Pantic. FP - Age : Leveraging Face Parsing Attention for Facial Age Estimation in the Wild , 2022. arXiv:2106.11145 [cs]

  23. [31]

    BlendGAN : Implicitly GAN Blending for Arbitrary Stylized Face Generation , 2021

    Mingcong Liu, Qiang Li, Zekui Qin, Guoxin Zhang, Pengfei Wan, and Wen Zheng. BlendGAN : Implicitly GAN Blending for Arbitrary Stylized Face Generation , 2021. arXiv:2110.11728 [cs]

  24. [32]

    Countering Language Drift with Seeded Iterated Learning

    Yuchen Lu, Soumye Singhal, Florian Strub, Aaron Courville, and Olivier Pietquin. Countering Language Drift with Seeded Iterated Learning . In Proceedings of the 37th International Conference on Machine Learning , pages 6437--6447. PMLR, 2020. ISSN: 2640-3498

  25. [33]

    Re- Aging GAN : Toward Personalized Face Age Transformation

    Farkhod Makhmudkhujaev, Sungeun Hong, and In Kyu Park. Re- Aging GAN : Toward Personalized Face Age Transformation . In 2021 IEEE / CVF International Conference on Computer Vision ( ICCV ) , pages 3888--3897, Montreal, QC, Canada, 2021. IEEE. TLDR: This work proposes Re-Aging ...

  26. [34]

    Changes in the Facial Skeleton With Aging : Implications and Clinical Applications in Facial Rejuvenation

    Bryan Mendelson and Chin-Ho Wong. Changes in the Facial Skeleton With Aging : Implications and Clinical Applications in Facial Rejuvenation . Aesthetic Plastic Surgery, 36 0 (4): 0 753--760, 2012

  27. [35]

    Null-text Inversion for Editing Real Images using Guided Diffusion Models , 2022

    Ron Mokady, Amir Hertz, Kfir Aberman, Yael Pritch, and Daniel Cohen-Or. Null-text Inversion for Editing Real Images using Guided Diffusion Models , 2022. arXiv:2211.09794 [cs]

  28. [36]

    Video Face Re - Aging : Toward Temporally Consistent Face Re - Aging , 2023

    Abdul Muqeet, Kyuchul Lee, Bumsoo Kim, Yohan Hong, Hyungrae Lee, Woonggon Kim, and KwangHee Lee. Video Face Re - Aging : Toward Temporally Consistent Face Re - Aging , 2023. arXiv:2311.11642 [cs] TLDR: A novel synthetic video dataset that features subjects across a diverse ran...

  29. [37]

    MyStyle : A Personalized Generative Prior

    Yotam Nitzan, Kfir Aberman, Qiurui He, Orly Liba, Michal Yarom, Yossi Gandelsman, Inbar Mosseri, Yael Pritch, and Daniel Cohen-Or. MyStyle : A Personalized Generative Prior . ACM Transactions on Graphics, 41 0 (6): 0 206:1--206:10, 2022 a . TLDR: This work introduces MyStyle, ...

  30. [38]

    LARGE : Latent - Based Regression through GAN Semantics

    Yotam Nitzan, Rinon Gal, Ofir Brenner, and Daniel Cohen-Or. LARGE : Latent - Based Regression through GAN Semantics . In 2022 IEEE / CVF Conference on Computer Vision and Pattern Recognition ( CVPR ) , pages 19217--19227, New Orleans, LA, USA, 2022 b . IEEE. TLDR: This work pr...

  31. [39]

    Domain Expansion of Image Generators , 2023

    Yotam Nitzan, Michaël Gharbi, Richard Zhang, Taesung Park, Jun-Yan Zhu, Daniel Cohen-Or, and Eli Shechtman. Domain Expansion of Image Generators , 2023. arXiv:2301.05225 [cs]

  32. [40]

    Lifespan Age Transformation Synthesis , 2020

    Roy Or-El, Soumyadip Sengupta, Ohad Fried, Eli Shechtman, and Ira Kemelmacher-Shlizerman. Lifespan Age Transformation Synthesis , 2020. arXiv:2003.09764 [cs]

  33. [41]

    StyleCLIP : Text - Driven Manipulation of StyleGAN Imagery , 2021

    Or Patashnik, Zongze Wu, Eli Shechtman, Daniel Cohen-Or, and Dani Lischinski. StyleCLIP : Text - Driven Manipulation of StyleGAN Imagery , 2021. arXiv:2103.17249 [cs]

  34. [42]

    Wang, Shengze Wang, and Roni Sengupta

    Luchao Qi, Jiaye Wu, Annie N. Wang, Shengze Wang, and Roni Sengupta. My3DGen : A Scalable Personalized 3D Generative Model , 2023. arXiv:2307.05468 [cs]

  35. [43]

    Learning Transferable Visual Models From Natural Language Supervision , 2021

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning Transferable Visual Models From Natural Language Supervision , 2021. arXiv:2103.00020 [cs]

  36. [44]

    Reinforced Disentanglement for Face Swapping without Skip Connection , 2023

    Xiaohang Ren, Xingyu Chen, Pengfei Yao, Heung-Yeung Shum, and Baoyuan Wang. Reinforced Disentanglement for Face Swapping without Skip Connection , 2023. arXiv:2307.07928 [cs]

  37. [45]

    Ricanek and T

    K. Ricanek and T. Tesafaye. MORPH : a longitudinal image database of normal adult age-progression. In 7th International Conference on Automatic Face and Gesture Recognition ( FGR06 ) , pages 341--345, 2006

  38. [46]

    Encoding in Style : a StyleGAN Encoder for Image -to- Image Translation

    Elad Richardson, Yuval Alaluf, Or Patashnik, Yotam Nitzan, Yaniv Azar, Stav Shapiro, and Daniel Cohen-Or. Encoding in Style : a StyleGAN Encoder for Image -to- Image Translation . In 2021 IEEE / CVF Conference on Computer Vision and Pattern Recognition ( CVPR ) , pages 2287--2...

  39. [47]

    Bermano, and Daniel Cohen-Or

    Daniel Roich, Ron Mokady, Amit H. Bermano, and Daniel Cohen-Or. Pivotal Tuning for Latent -based Editing of Real Images , 2021. arXiv:2106.05744 [cs]

  40. [48]

    DEX : Deep EXpectation of Apparent Age from a Single Image

    Rasmus Rothe, Radu Timofte, and Luc Van Gool. DEX : Deep EXpectation of Apparent Age from a Single Image . In 2015 IEEE International Conference on Computer Vision Workshop ( ICCVW ) , pages 252--257, Santiago, Chile, 2015. IEEE. TLDR: The proposed method, Deep EXpectation (DE...

  41. [49]

    Semantic Image Inversion and Editing using Rectified Stochastic Differential Equations , 2024

    Litu Rout, Yujia Chen, Nataniel Ruiz, Constantine Caramanis, Sanjay Shakkottai, and Wen-Sheng Chu. Semantic Image Inversion and Editing using Rectified Stochastic Differential Equations , 2024. arXiv:2410.10792

  42. [50]

    DreamBooth : Fine Tuning Text -to- Image Diffusion Models for Subject - Driven Generation , 2023

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Yael Pritch, Michael Rubinstein, and Kfir Aberman. DreamBooth : Fine Tuning Text -to- Image Diffusion Models for Subject - Driven Generation , 2023. arXiv:2208.12242 [cs]

  43. [51]

    InterFaceGAN : Interpreting the Disentangled Face Representation Learned by GANs , 2020

    Yujun Shen, Ceyuan Yang, Xiaoou Tang, and Bolei Zhou. InterFaceGAN : Interpreting the Disentangled Face Representation Learned by GANs , 2020. arXiv:2005.09635 [cs, eess]

  44. [52]

    A Compositional and Dynamic Model for Face Aging

    Jinli Suo, Song-Chun Zhu, Shiguang Shan, and Xilin Chen. A Compositional and Dynamic Model for Face Aging . IEEE Transactions on Pattern Analysis and Machine Intelligence, 32 0 (3): 0 385--401, 2010. Conference Name: IEEE Transactions on Pattern Analysis and Machine Intelligence

  45. [53]

    Inside Out

    Arthur Swift, Steven Liew, Susan Weinkle, Julie K Garcia, and Michael B Silberberg. The Facial Aging Process From the “ Inside Out ”. Aesthetic Surgery Journal, 41 0 (10): 0 1107--1119, 2020

  46. [54]

    Face Aging with Identity - Preserved Conditional Generative Adversarial Networks

    Xu Tang, Zongwei Wang, Weixin Luo, and Shenghua Gao. Face Aging with Identity - Preserved Conditional Generative Adversarial Networks . In 2018 IEEE / CVF Conference on Computer Vision and Pattern Recognition , pages 7939--7947, Salt Lake City, UT, USA, 2018. IEEE

  47. [55]

    Facial aging simulator considering geometry and patch-tiled texture

    Yusuke Tazoe, Hiroaki Gohara, Akinobu Maejima, and Shigeo Morishima. Facial aging simulator considering geometry and patch-tiled texture. In ACM SIGGRAPH 2012 Posters , page 1, New York, NY, USA, 2012. Association for Computing Machinery

  48. [56]

    Exploring 3D -aware Lifespan Face Aging via Disentangled Shape - Texture Representations , 2023

    Qianrui Teng, Rui Wang, Xing Cui, Peipei Li, and Zhaofeng He. Exploring 3D -aware Lifespan Face Aging via Disentangled Shape - Texture Representations , 2023. arXiv:2312.16881 [cs]

  49. [57]

    Tiddeman, M

    B. Tiddeman, M. Burt, and D. Perrett. Prototyping and transforming facial textures for perception research. IEEE Computer Graphics and Applications, 21 0 (5): 0 42--50, 2001. Conference Name: IEEE Computer Graphics and Applications

  50. [58]

    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, 40 0 (4): 0 133:1--133:14, 2021

  51. [59]

    The Irishman : Robert De Niro Recreated Iconic GoodFellas Scene To Test De - Ageing Tech – Exclusive , 2019

    Ben Travis. The Irishman : Robert De Niro Recreated Iconic GoodFellas Scene To Test De - Ageing Tech – Exclusive , 2019

  52. [60]

    InstantID : Zero -shot Identity - Preserving Generation in Seconds , 2024

    Qixun Wang, Xu Bai, Haofan Wang, Zekui Qin, and Anthony Chen. InstantID : Zero -shot Identity - Preserving Generation in Seconds , 2024. arXiv:2401.07519 [cs]

  53. [61]

    Towards Real - World Blind Face Restoration with Generative Facial Prior , 2021

    Xintao Wang, Yu Li, Honglun Zhang, and Ying Shan. Towards Real - World Blind Face Restoration with Generative Facial Prior , 2021. arXiv:2101.04061 [cs]

  54. [62]

    GAN Inversion : A Survey , 2022

    Weihao Xia, Yulun Zhang, Yujiu Yang, Jing-Hao Xue, Bolei Zhou, and Ming-Hsuan Yang. GAN Inversion : A Survey , 2022. arXiv:2101.05278 [cs]

  55. [63]

    Diverse and Lifespan Facial Age Transformation Synthesis with Identity Variation Rationality Metric , 2024

    Jiu-Cheng Xie, Jun Yang, Wenqing Wang, Feng Xu, and Hao Gao. Diverse and Lifespan Facial Age Transformation Synthesis with Identity Variation Rationality Metric , 2024. arXiv:2401.14036 [cs]

  56. [64]

    Designing One Unified Framework for High - Fidelity Face Reenactment and Swapping

    Chao Xu, Jiangning Zhang, Yue Han, Guanzhong Tian, Xianfang Zeng, Ying Tai, Yabiao Wang, Chengjie Wang, and Yong Liu. Designing One Unified Framework for High - Fidelity Face Reenactment and Swapping . In Computer Vision – ECCV 2022 , pages 54--71, Cham, 2022 a . Springer Natu...

  57. [65]

    MobileFaceSwap : A Lightweight Framework for Video Face Swapping , 2022 b

    Zhiliang Xu, Zhibin Hong, Changxing Ding, Zhen Zhu, Junyu Han, Jingtuo Liu, and Errui Ding. MobileFaceSwap : A Lightweight Framework for Video Face Swapping , 2022 b . arXiv:2201.03808

  58. [66]

    High Resolution Face Age Editing , 2020

    Xu Yao, Gilles Puy, Alasdair Newson, Yann Gousseau, and Pierre Hellier. High Resolution Face Age Editing , 2020. arXiv:2005.04410 [cs]

  59. [67]

    MyStyle ++: A Controllable Personalized Generative Prior , 2023

    Libing Zeng, Lele Chen, Yi Xu, and Nima Kalantari. MyStyle ++: A Controllable Personalized Generative Prior , 2023. arXiv:2306.04865 [cs]

  60. [68]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In CVPR, 2018

  61. [69]

    Age Progression / Regression by Conditional Adversarial Autoencoder , 2017

    Zhifei Zhang, Yang Song, and Hairong Qi. Age Progression / Regression by Conditional Adversarial Autoencoder , 2017. arXiv:1702.08423 [cs]

  62. [70]

    InverseMeetInsert : Robust Real Image Editing via Geometric Accumulation Inversion in Guided Diffusion Models , 2024

    Yan Zheng and Lemeng Wu. InverseMeetInsert : Robust Real Image Editing via Geometric Accumulation Inversion in Guided Diffusion Models , 2024. arXiv:2409.11734 version: 1 TLDR: GEO is introduced, an exceptionally versatile image editing technique designed to cater to customize...

  63. [71]

    Jun-Yan Zhu, Taesung Park, Phillip Isola, and Alexei A. Efros. Unpaired image-to-image translation using cycle-consistent adversarial networks. In 2017 IEEE International Conference on Computer Vision (ICCV), pages 2242--2251, 2017

  64. [72]

    Production- Ready Face Re - Aging for Visual Effects

    Gaspard Zoss, Prashanth Chandran, Eftychios Sifakis, Markus Gross, Paulo Gotardo, and Derek Bradley. Production- Ready Face Re - Aging for Visual Effects . ACM Transactions on Graphics, 41 0 (6): 0 1--12, 2022. TLDR: This paper presents the first practical, fully-automatic and...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.