FCR: Investigating Generative AI models for Forensic Craniofacial Reconstruction

Dinesh Singh; Ravi Shankar Prasad

REVIEW 4 major objections 5 minor 1 cited by

FCR: Investigating Generative AI models for Forensic Craniofacial Reconstruction

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A generative model can turn 2D skull X-rays into recognizable candidate faces, the paper argues.

desk verdict A genuinely new application of GANs to 2D X-ray craniofacial reconstruction, but the evaluation leaks identity across train/test, so the headline forensic-identification claim is not supported as stated. read the letter →

arxiv 2508.18031 v2 pith:TQHOJOGL submitted 2025-08-25 cs.CV

Ravi Shankar Prasad , Dinesh Singh This is my paper · ORCID

classification cs.CV

keywords craniofacialreconstructionforensicidentificationgenerativeadversarialnetworksimage-to-imagetranslationskull-to-facesynthesis2DX-rayfaceretrievalcontrastivelearning

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 generative models can reconstruct a face from a 2D X-ray of the skull, and that the best of them, FastCUT, preserves enough identity to act as an assisting tool in forensic craniofacial identification. This matters because standard forensic reconstruction relies on clay sculpting by experts or costly CT-based pipelines, whereas 2D X-rays are cheap and widely available. The paper reports that FastCUT beats CycleGAN, cGAN, and CUT on FID, IS, and ArcFace similarity, though cGAN is closer on perceptual similarity.

What carries the argument

The load-bearing mechanism is FastCUT, a one-sided unpaired image-to-image translation model that combines an adversarial loss with a patch-wise contrastive (NCE) loss, matching each local patch of the generated face to the corresponding patch of the input skull while pushing other patches apart, plus an identity regularizer that penalizes changes to images already in the face domain. The pipeline also depends on a preprocessing step that manually removes soft tissue from X-rays so the input resembles a skull, and on the S2F paired dataset of frontal and lateral views. This machinery is what lets the model transfer structural details from skull to face without paired supervision.

What would settle it

Redo the training/test split so that all views of each person are kept together, then compare test ArcFace similarity and retrieval recall; if the scores fall to near chance, the current claims are inflated by identity leakage.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that skull-to-face translation can be done from ordinary 2D X-ray images rather than 3D CT data, and that the resulting faces are realistic and identifiable enough for forensic candidate retrieval. Using an extended S2F dataset of 51 individuals with frontal and lateral skull-face pairs, the authors compare four generative models and find FastCUT best: FID 63.65, SSIM 0.66, ArcFace similarity 0.5167. In retrieval tests against galleries of 100 and 200 real faces, correct identities appear in the top-k at high recall but with low mean average precision, so the authors position the framework as a candidate-shortlisting tool rather than a final identification system.

Load-bearing premise

The results assume that a random 80:20 split of the 102 skull-face pairs keeps test identities unseen, but because each of the 51 people contributes both a frontal and a lateral pair, the same person can appear in both training and test, so the identity scores may measure memorization rather than generalization.

Editorial extensions

If this is right

If FastCUT's results hold, forensic teams could generate candidate faces from 2D skull X-rays in hours instead of weeks.
The same framework could be applied to lateral and frontal views separately, and the better-performing view could be used for retrieval.
The high recall but low mAP implies the system can narrow a missing-person database but should not be used alone for positive identification.
The comparison suggests lightweight one-sided models are more suitable than heavy cycle-consistency models for skull-to-face translation under limited data.

Reading between the lines

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

The reported identity-preservation numbers likely overstate generalization because a random pair split may put the same person's frontal and lateral views in both training and test; an identity-disjoint split is the natural next check.
If identity-disjoint evaluation confirms the result, 2D X-ray reconstruction could serve as a low-cost first-pass screening step to narrow missing-person databases before expensive DNA or CT analysis.
The high-recall/low-mAP pattern suggests the method is suited to candidate shortlisting rather than final identification, and retrieval ranking could be improved by training the generator with an explicit face-recognition loss.
Cross-population transfer remains untested because all subjects are young adults from one population; applying the model to skull X-rays from other age and ancestry groups would reveal whether the skull-to-face mapping generalizes.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The short version: this paper does something new and worth examining—first GAN-based skull-to-face translation from 2D X-rays, plus a modest extension of the authors' own S2F dataset—but the central quantitative evidence for the forensic claim is undermined by a train/test split that does not separate identities. I would send it to peer review with strong encouragement for revision, not desk reject it.

What is actually useful: the task itself is legitimate and underexplored. Prior generative craniofacial work used CT or 3D data; 2D X-rays are cheaper and more accessible, so the application has practical value. The paper compares four standard GAN baselines (CycleGAN, cGAN, CUT, FastCUT), adds a retrieval module with four backbone networks, and extends the S2F dataset from 40 to 51 individuals. The authors also include failure cases and honestly acknowledge that FID and IS are unstable with 20 test samples. That is a reasonable proof-of-concept structure.

The soft spot is not minor. The dataset has 102 image pairs—each of the 51 subjects contributes a frontal and a lateral pair. The split in Section 3 is described as randomly selecting 20 pairs for testing, with no mention of an identity-disjoint split. With 20 test pairs out of 102, most test identities will also have their other view in the training set. For unpaired models like CycleGAN, CUT, and FastCUT, the face domain contains the same person's face during training, so the generator can learn to reproduce that identity's appearance rather than generalize from skull structure. ArcFace similarity of 0.52 and the high recall@k values likely reflect this memorization. The retrieval evaluation adds another problem: the gallery of 200 faces is from the same institution, but the paper never states whether the 51 volunteers overlap with the gallery, and there is no chance baseline or comparison against direct skull-to-face matching. With mAP below 10 in essentially all settings, the retrieval claim is weak even aside from the split issue.

To be fair, the paper is transparent about dataset size, demographic homogeneity, and lack of an external benchmark. The manual soft-tissue removal is described carefully, though not independently validated. But the identity leakage is a load-bearing flaw: the conclusion that these models can be an assisting tool for forensic identification rests almost entirely on identity-preservation and retrieval numbers that may simply reflect training identities.

Who gets value from this? Researchers in forensic imaging or cross-modal translation might use the dataset extension or the baseline comparisons, but only after the evaluation is redone. The paper needs an identity-disjoint split, a specified gallery with no training overlap, chance-level baselines, and preferably more test samples. As it stands, the evidence supports feasibility at best, not forensic utility.

My recommendation: send to peer review, because the topic is real and the flaws are fixable. But a serious referee should require the re-evaluation before publication.

Referee Report

4 major / 5 minor

Summary. The paper proposes a framework for craniofacial reconstruction from 2D X-ray skull images using four generative models (CycleGAN, cGAN, CUT, FastCUT). The authors extend the S2F dataset to 51 individuals (102 frontal/lateral image pairs), manually remove soft tissue from the X-ray images, train the generative models, and evaluate the generated faces using FID, IS, SSIM, LPIPS, and ArcFace similarity. They also propose a retrieval framework in which a generated face is used as a query against a gallery of real faces, reporting Recall@k and mAP@k for several backbones and gallery sizes. The paper concludes that FastCUT performs best and that these generative models can serve as an assisting tool for forensic craniofacial identification.

Significance. If the result held, the approach would be valuable because 2D X-rays are cheaper and more accessible than CT scans for forensic skull analysis. The paper's strengths include the construction of a paired skull–face dataset with ethical approval, a systematic comparison of four GAN-based methods, and a practical retrieval formulation with several feature extractors. However, the evaluation has a load-bearing flaw: the train/test split is by image pairs rather than by identity, so test identities almost certainly appear in training. In addition, the retrieval gallery is not specified to be disjoint from the dataset subjects, and no baseline is reported against direct skull-to-face matching. The central claim of forensic utility is therefore not currently supported.

major comments (4)

[Section 3, dataset split] The dataset is split into an 80:20 ratio, with 20 of 102 pairs randomly selected for testing. Because each of the 51 subjects contributes both a frontal and a lateral pair, this pair-level split almost certainly places the other view of a test identity in the training set. The reported ArcFace similarity (e.g., FastCUT: 0.5167) and Recall@k values in Table 2b may therefore measure the models' ability to reproduce identity-specific appearance seen during training rather than generalization to unseen skulls. Please re-run all experiments with an identity-disjoint split (both views of a subject in the same fold) and report the same metrics. This is the central issue for the conclusion in Section 6.1.
[Section 4.6, retrieval gallery] The face gallery is described only as 'collected from IITMandi student volunteers,' with no statement about whether it overlaps with the 51 dataset subjects. If the gallery contains face images of the identities used for testing (or of identities whose images appear in training), the retrieval results in Table 2b are inflated and cannot support the forensic identification claim. The authors must specify the gallery composition, the number of unique identities, and the degree of overlap with the training/test set, and should ideally evaluate with a public face database (e.g., LFW) for external validation.
[Section 6.1, conclusion vs. baselines] The conclusion states that 'these generative models can be used as an assisting tool for craniofacial identifications,' but no baseline comparison is provided. The paper should compare the proposed generative-retrieval pipeline against (a) direct skull-to-face matching using the cross-domain embedding method of ref. 13, (b) a simple nearest-neighbor baseline on skull image features, and (c) retrieval using the ground-truth face images as queries to calibrate an upper bound. Without these baselines, the added value of the generative reconstruction step is not demonstrated.
[Sections 4.2 and 5.2, validation of inputs and metric stability] The manual soft-tissue removal procedure (Section 4.2) is not validated; if the masks do not accurately isolate skull structure, the 'skull' domain is ill-defined and all downstream results are questionable. Additionally, the authors acknowledge in Section 5.2 that FID and IS are 'not statistically stable' for 20 test samples, yet these metrics are used to rank methods. Please provide validation of the soft-tissue removal (e.g., inter-annotator agreement or comparison with CT-derived skulls) and report confidence intervals or bootstrap estimates for FID/IS, or relegate them to secondary consideration.

minor comments (5)

[Section 4.3 and Figure 4] The textual description of generators G_X and G_Y appears inconsistent with the equations and Figure 4: the text describes G_X as having a pretrained ResNet18 encoder and G_Y as fully trainable, but the loss definitions and figure assign G_Y to skull-to-face translation and G_X to face-to-skull translation. Please reconcile the notation.
[Section 4.5, Eq. (11)] The statement that FastCUT corresponds to lambda_X=10 and lambda_Y=0 should be checked against the original CUT paper; as written, it appears to contradict the standard FastCUT configuration (lambda_X=10, lambda_Y=0 in the original, but the surrounding text says lambda_X and lambda_Y are 1).
[Table 2b] The repeated 'Recall@k mAP@k' headers for each gallery size make the table difficult to parse; please use multi-level column headers or split the table into separate subtables for gallery sizes 100 and 200.
[Section 6.1] The phrase 'within the tok-k' is a typo and should read 'within the top-k'.
[Section 7, Data availability] The data availability statement says 'Our data will be made available on request,' but the face images used for training and testing are not shared due to privacy. Please clarify precisely which data (e.g., preprocessed masks, generated images, feature vectors) will be shared, as this affects reproducibility.

Circularity Check

1 steps flagged · score 6.0 of 10

Random pair-level 80:20 split leaks identities into training, so FastCUT's ArcFace and retrieval scores can reflect memorization of seen identities rather than generalization to unseen skulls.

fitted input called prediction [Section 3 (Dataset) and Section 5.2 (Experimental results), Table 2]
"Initially, we have 102 skull and face pairs, which include 51 lateral and 51 frontal views, respectively. The dataset is split into an 80:20 ratio for training and testing. Out of 102 pairs, 20 pairs are randomly selected for testing. ... Additionally, FastCUT attains the highest ArcFace similarity, suggesting improved identity preservation when compared to other generative models."

Because each of the 51 identities contributes both a frontal and a lateral pair, a random split of 102 pairs into 82 training and 20 test pairs does not keep identities disjoint. In expectation roughly 16 of the 20 test pairs belong to identities whose other pair is in the training set. The identity-preservation and retrieval metrics reported in Section 5.2 (ArcFace similarity, recall@k, mAP@k) are therefore evaluated on faces whose identity-specific appearance was available during training; the generator can memorize that appearance rather than predict an unseen person's face from a skull. The conclusion that FastCUT can act as a forensic identification tool rests on these contaminated metrics, so the measured 'prediction' is partly a fitted reproduction of training identities.

full rationale

The paper is an empirical evaluation study rather than a formal derivation, so most circularity patterns do not apply. The main load-bearing problem is the evaluation protocol: the random pair-level split combined with two pairs per identity means test identities substantially overlap with training identities, so the identity-sensitive results (ArcFace similarity, retrieval recall) are not a valid test of generalization to previously unseen people. This is a partial circularity in the sense that the reported identity-preservation 'prediction' can be achieved by memorizing individuals already seen during training. The paper itself acknowledges that FID/IS are unstable with only 20 test samples, but it does not address the more serious identity-overlap confound. The self-citation of the S2F dataset from the authors' prior work is notable but not itself the load-bearing circular step, since the data collection and preprocessing are described in this paper. No uniqueness-theorem import or ansatz-smuggling pattern is present. The central forensic-assistance claim is thus only partially supported: quality metrics like FID and qualitative examples retain some independent content, but the identification claim relies on evaluation that reduces by construction to training-set identity appearance.

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

The central evaluation rests on a self-created, self-annotated dataset (S2F, extended from the authors' prior work), a manual soft-tissue removal step without external validation, and an internal retrieval gallery. No independent benchmark is used, so the conclusions are only as strong as these internal assumptions.

free parameters (7)

CycleGAN cycle consistency weight (lambda_cycle) = 10
Set to 10 following the original CycleGAN paper; hand-chosen, not fitted to the S2F dataset.
CycleGAN identity weight (lambda_id) = 5
Set to 5 per CycleGAN; affects the trade-off between identity preservation and translation quality.
CUT temperature (tau) = 0.07
Standard value from CUT paper; controls the sharpness of the patch-wise contrastive distribution.
CUT/FastCUT loss weights (lambda_X, lambda_Y) = 1 and 1 (FastCUT: 10 and 0)
Default values from the CUT paper; FastCUT is instantiated by setting lambda_X=10 and lambda_Y=0.
Learning rate = 0.0001
Chosen by authors for all models; not reported as tuned.
Batch size = 16
Selected for training on an RTX A5000 GPU; standard choice.
Training epochs = 500
All models trained for 500 epochs; no early stopping described.

assumptions (5)

domain assumption After manual soft tissue removal, the 2D X-ray image faithfully represents the skull's hard tissue and retains identity-relevant structure.
Section 4.2 describes polygon-based segmentation to mask out soft tissue; this is a manual, subjective process with no validation against CT ground truth.
domain assumption ArcFace similarity between the generated face and the real face is a valid measure of identity preservation for forensic identification.
Used in Section 5.2 as the identity metric; assumes the pre-trained ArcFace model generalizes to this cross-modal, small-sample setting.
domain assumption The random 80:20 split of image pairs creates statistically independent train and test identities.
Section 3 splits 102 pairs randomly; because each individual contributes two views (frontal and lateral), an individual may appear in both training and test sets, violating independence.
domain assumption The retrieval gallery contains the ground-truth face corresponding to each query generated face.
Recall@k and mAP are only meaningful if the correct identity is present in the gallery; the paper never states this explicitly.
domain assumption Cycle-consistency and identity losses are appropriate inductive biases for skull-to-face translation.
Section 4.3/4.5 borrow these losses from image translation literature; their suitability for high-stakes forensic reconstruction is assumed.

how reviews work

0 comments

Cite this review

Pith. "Pith review of FCR: Investigating Generative AI models for Forensic Craniofacial Reconstruction." pith.science (2026). https://pith.science/paper/TQHOJOGL

@misc{pith2026250818031,
  author       = {Pith},
  title        = {Pith review of: FCR: Investigating Generative AI models for Forensic Craniofacial Reconstruction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TQHOJOGL}},
  note         = {Machine review of arXiv:2508.18031}
}

read the original abstract

Craniofacial reconstruction in forensics is one of the processes to identify victims of crime and natural disasters. Identifying an individual from their remains plays a crucial role when all other identification methods fail. Traditional methods for this task, such as clay-based craniofacial reconstruction, require expert domain knowledge and are a time-consuming process. At the same time, other probabilistic generative models like the statistical shape model or the Basel face model fail to capture the skull and face cross-domain attributes. Looking at these limitations, we propose a generic framework for craniofacial reconstruction from 2D X-ray images. Here, we used various generative models (i.e., CycleGANs, cGANs, etc) and fine-tune the generator and discriminator parts to generate more realistic images in two distinct domains, which are the skull and face of an individual. This is the first time where 2D X-rays are being used as a representation of the skull by generative models for craniofacial reconstruction. We have evaluated the quality of generated faces using FID, IS, and SSIM scores. Finally, we have proposed a retrieval framework where the query is the generated face image and the gallery is the database of real faces. By experimental results, we have found that these generative models can be used as an assisting tool for craniofacial identifications in forensic science.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

Cranio-Diff: Diffusion-based Cross-domain Craniofacial Reconstruction with 2D X-ray Skull Guidance and Structural Identity Constraints
cs.CV 2026-06 unverdicted novelty 6.0 of 10

Cranio-Diff applies a conditioned diffusion model with ControlNet and text prompts to reconstruct faces from skull X-rays, evaluated on a synthesized dataset of 4320 samples from 120 subjects across age and BMI variations.

Reference graph

Works this paper leans on

42 extracted references · 33 canonical work pages · cited by 1 Pith paper

[1]

forensic sciences57, 1442–1447 (2012)

Hwang, H.-S.et al.Facial soft tissue thickness database for craniofacial reconstruction in korean adults.J. forensic sciences57, 1442–1447 (2012). 2.Wilkinson, C. Facial reconstruction–anatomical art or artistic anatomy?J. anatomy216, 235–250 (2010)

work page 2012
[3]

Damas, S.et al.Forensic identification by computer-aided craniofacial superimposition: a survey.ACM Comput. Surv. (CSUR)43, 1–27 (2011)

work page 2011
[4]

R.et al.Computer vision and soft computing for automatic skull–face overlay in craniofacial superimposition.F orensic Sci

Campomanes-Alvarez, B. R.et al.Computer vision and soft computing for automatic skull–face overlay in craniofacial superimposition.F orensic Sci. Int.245, 77–86 (2014)

work page 2014
[5]

Claes, P.et al.Computerized craniofacial reconstruction: conceptual framework and review.F orensic science international 201, 138–145 (2010)

work page 2010
[6]

& Duncan, C

Dai, H., Pears, N., Smith, W. & Duncan, C. Statistical modeling of craniofacial shape and texture.Int. J. Comput. Vis.128, 547–571 (2020). 7.Goodfellow, I.et al.Generative adversarial networks.Commun. ACM63, 139–144 (2020)

work page 2020
[8]

& Efros, A

Zhu, J.-Y ., Park, T., Isola, P. & Efros, A. A. Unpaired image-to-image translation using cycle-consistent adversarial networks. InProceedings of the IEEE international conference on computer vision, 2223–2232 (2017)

2017
[9]

A., Zhang, R

Park, T., Efros, A. A., Zhang, R. & Zhu, J.-Y . Contrastive learning for unpaired image-to-image translation. InComputer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part IX 16, 319–345 (Springer, 2020)

work page 2020
[10]

& Duan, F

Li, Y ., Chang, L., Qiao, X., Liu, R. & Duan, F. Craniofacial reconstruction based on least square support vector regression. In2014 IEEE international conference on systems, man, and cybernetics (SMC), 1147–1151 (IEEE, 2014)

work page 2014

Show all 42 references

[11]

M., Glaunès, J

Berar, M., Tilotta, F. M., Glaunès, J. A. & Rozenholc, Y . Craniofacial reconstruction as a prediction problem using a latent root regression model.F orensic science international210, 228–236 (2011)

2011
[12]

& Graph.97, 258–267 (2021)

Jia, B.et al.Craniofacial reconstruction based on heat flow geodesic grid regression (hf-ggr) model.Comput. & Graph.97, 258–267 (2021)

2021
[13]

Prasad, R. S. & Singh, D. Cross-domain identity representation for skull to face matching with benchmark dataset.arXiv preprint arXiv:2507.08329(2025)

2025 arXiv
[14]

In2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 462–469 (IEEE, 2021)

Pengyue, L.et al.Cfr-gan: a generative model for craniofacial reconstruction. In2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 462–469 (IEEE, 2021). 13/15

2021
[15]

Li, Y .et al.Cr-gan: Automatic craniofacial reconstruction for personal identification.Pattern Recognit.124, 108400 (2022)

2022
[16]

InProceedings of the 30th ACM international conference on multimedia, 759–768 (2022)

Zhang, N.et al.An end-to-end conditional generative adversarial network based on depth map for 3d craniofacial reconstruction. InProceedings of the 30th ACM international conference on multimedia, 759–768 (2022)

2022
[17]

InSIGGRAPH Asia 2024 Technical Communications, 1–4 (2024)

Zhao, J.et al.Intrinsic morphological relationship guided 3d craniofacial reconstruction using siamese cycle attention gan. InSIGGRAPH Asia 2024 Technical Communications, 1–4 (2024)

2024
[18]

InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 9307–9315 (2024)

Jayasumana, S.et al.Rethinking fid: Towards a better evaluation metric for image generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 9307–9315 (2024). 19.Barratt, S. & Sharma, R. A note on the inception score.arXiv preprint arXiv:18...

2024 arXiv
[21]

A., Shechtman, E

Zhang, R., Isola, P., Efros, A. A., Shechtman, E. & Wang, O. The unreasonable effectiveness of deep features as a perceptual metric. InProceedings of the IEEE conference on computer vision and pattern recognition, 586–595 (2018)

2018
[22]

& Zafeiriou, S

Deng, J., Guo, J., Xue, N. & Zafeiriou, S. Arcface: Additive angular margin loss for deep face recognition. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 4690–4699 (2019). 23.Goodfellow, I. J.et al.Generative adversarial nets.Adv. neural info...

2019
[24]

& Efros, A

Isola, P., Zhu, J.-Y ., Zhou, T. & Efros, A. A. Image-to-image translation with conditional adversarial networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, 1125–1134 (2017)

2017
[25]

& Gong, M

Yi, Z., Zhang, H., Tan, P. & Gong, M. Dualgan: Unsupervised dual learning for image-to-image translation. InProceedings of the IEEE international conference on computer vision, 2849–2857 (2017)

2017
[26]

Kim, T., Cha, M., Kim, H., Lee, J. K. & Kim, J. Learning to discover cross-domain relations with generative adversarial networks. InInternational conference on machine learning, 1857–1865 (Pmlr, 2017)

2017
[27]

& Kautz, J

Liu, M.-Y ., Breuel, T. & Kautz, J. Unsupervised image-to-image translation networks.Adv. neural information processing systems30(2017)

2017
[28]

& Kautz, J

Huang, X., Liu, M.-Y ., Belongie, S. & Kautz, J. Multimodal unsupervised image-to-image translation. InProceedings of the European conference on computer vision (ECCV), 172–189 (2018)

2018
[29]

In Proceedings of the IEEE conference on computer vision and pattern recognition, 8789–8797 (2018)

Choi, Y .et al.Stargan: Unified generative adversarial networks for multi-domain image-to-image translation. In Proceedings of the IEEE conference on computer vision and pattern recognition, 8789–8797 (2018)

2018
[30]

& Courville, A

Almahairi, A., Rajeshwar, S., Sordoni, A., Bachman, P. & Courville, A. Augmented cyclegan: Learning many-to-many mappings from unpaired data. InInternational conference on machine learning, 195–204 (PMLR, 2018). 31.Zhu, J.-Y .et al.Toward multimodal image-to-image translation....

2017
[32]

& Yang, M.-H

Lee, H.-Y ., Tseng, H.-Y ., Huang, J.-B., Singh, M. & Yang, M.-H. Diverse image-to-image translation via disentangled representations. InProceedings of the European conference on computer vision (ECCV), 35–51 (2018)

2018
[33]

InProceedings of the IEEE/CVF international conference on computer vision, 10551–10560 (2019)

Liu, M.-Y .et al.Few-shot unsupervised image-to-image translation. InProceedings of the IEEE/CVF international conference on computer vision, 10551–10560 (2019)

2019
[34]

& Yan, Y

Tang, H., Xu, D., Sebe, N. & Yan, Y . Attention-guided generative adversarial networks for unsupervised image-to-image translation. In2019 International Joint Conference on Neural Networks (IJCNN), 1–8 (IEEE, 2019). 35.Zhang, R., Pfister, T. & Li, J. Harmonic unpaired image-to...

2019 arXiv
[36]

Gokaslan, A., Ramanujan, V ., Ritchie, D., Kim, K. I. & Tompkin, J. Improving shape deformation in unsupervised image-to-image translation. InProceedings of the European Conference on Computer Vision (ECCV), 649–665 (2018)

2018
[37]

& Xing, E

Liang, X., Zhang, H., Lin, L. & Xing, E. Generative semantic manipulation with mask-contrasting gan. InProceedings of the European Conference on Computer Vision (ECCV), 558–573 (2018)

2018
[38]

& Loy, C

Wu, W., Cao, K., Li, C., Qian, C. & Loy, C. C. Transgaga: Geometry-aware unsupervised image-to-image translation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 8012–8021 (2019). 39.Dwyer, B., Nelson, J., Hansen, T.et al.Roboflow (2024). ...

2019
[40]

Bachman, P., Hjelm, R. D. & Buchwalter, W. Learning representations by maximizing mutual information across views. Adv. neural information processing systems32(2019). 14/15

2019
[41]

& Hinton, G

Chen, T., Kornblith, S., Norouzi, M. & Hinton, G. A simple framework for contrastive learning of visual representations. InInternational conference on machine learning, 1597–1607 (PmLR, 2020)

2020
[42]

Chen, X., Fan, H., Girshick, R. & He, K. Improved baselines with momentum contrastive learning.arXiv preprint arXiv:2003.04297(2020)

2020 arXiv
[43]

InProceedings of the IEEE/CVF international conference on computer vision, 12116–12125 (2021)

Zaeemzadeh, A.et al.Face image retrieval with attribute manipulation. InProceedings of the IEEE/CVF international conference on computer vision, 12116–12125 (2021)

2021
[44]

& Yan, S

Zhao, B., Feng, J., Wu, X. & Yan, S. Memory-augmented attribute manipulation networks for interactive fashion search. InProceedings of the IEEE conference on computer vision and pattern recognition, 1520–1528 (2017)

2017
[45]

& Kovashka, A

Yu, Z. & Kovashka, A. Syntharch: Interactive image search with attribute-conditioned synthesis. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops, 170–171 (2020). 46.Salimans, T.et al.Improved techniques for training gans.Adv. neural...

2020
[47]

& Sun, J

He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. InProceedings of the IEEE conference on computer vision and pattern recognition, 770–778 (2016)

2016
[48]

& Weinberger, K

Huang, G., Liu, Z., Van Der Maaten, L. & Weinberger, K. Q. Densely connected convolutional networks. InProceedings of the IEEE conference on computer vision and pattern recognition, 4700–4708 (2017)

2017
[49]

D., Raghavan, P

Manning, C. D., Raghavan, P. & Schütze, H.Introduction to Information Retrieval(Cambridge University Press, Cambridge, UK, 2009)

2009
[50]

& Hochreiter, S

Heusel, M., Ramsauer, H., Unterthiner, T., Nessler, B. & Hochreiter, S. Gans trained by a two time-scale update rule converge to a local nash equilibrium.Adv. neural information processing systems30(2017)

2017
[51]

C., Sheikh, H

Wang, Z., Bovik, A. C., Sheikh, H. R. & Simoncelli, E. P. Image quality assessment: from error visibility to structural similarity.IEEE transactions on image processing13, 600–612 (2004). 15/15

2004

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

forensic sciences57, 1442–1447 (2012)

Hwang, H.-S.et al.Facial soft tissue thickness database for craniofacial reconstruction in korean adults.J. forensic sciences57, 1442–1447 (2012). 2.Wilkinson, C. Facial reconstruction–anatomical art or artistic anatomy?J. anatomy216, 235–250 (2010)

work page 2012

[2] [3]

Damas, S.et al.Forensic identification by computer-aided craniofacial superimposition: a survey.ACM Comput. Surv. (CSUR)43, 1–27 (2011)

work page 2011

[3] [4]

R.et al.Computer vision and soft computing for automatic skull–face overlay in craniofacial superimposition.F orensic Sci

Campomanes-Alvarez, B. R.et al.Computer vision and soft computing for automatic skull–face overlay in craniofacial superimposition.F orensic Sci. Int.245, 77–86 (2014)

work page 2014

[4] [5]

Claes, P.et al.Computerized craniofacial reconstruction: conceptual framework and review.F orensic science international 201, 138–145 (2010)

work page 2010

[5] [6]

& Duncan, C

Dai, H., Pears, N., Smith, W. & Duncan, C. Statistical modeling of craniofacial shape and texture.Int. J. Comput. Vis.128, 547–571 (2020). 7.Goodfellow, I.et al.Generative adversarial networks.Commun. ACM63, 139–144 (2020)

work page 2020

[6] [8]

& Efros, A

Zhu, J.-Y ., Park, T., Isola, P. & Efros, A. A. Unpaired image-to-image translation using cycle-consistent adversarial networks. InProceedings of the IEEE international conference on computer vision, 2223–2232 (2017)

2017

[7] [9]

A., Zhang, R

Park, T., Efros, A. A., Zhang, R. & Zhu, J.-Y . Contrastive learning for unpaired image-to-image translation. InComputer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part IX 16, 319–345 (Springer, 2020)

work page 2020

[8] [10]

& Duan, F

Li, Y ., Chang, L., Qiao, X., Liu, R. & Duan, F. Craniofacial reconstruction based on least square support vector regression. In2014 IEEE international conference on systems, man, and cybernetics (SMC), 1147–1151 (IEEE, 2014)

work page 2014