Beyond Photo Realism for Domain Adaptation from Synthetic Data

Connor DeFanti; Jonathan Tompson; Ken Perlin; Kristofer Schlachter; Sebastian Herscher

REVIEW 2 major objections 6 minor 28 references

Beyond Photo Realism for Domain Adaptation from Synthetic Data

T0 review · 2 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read This paper claims that an ensemble of GAN-refined synthetic images trains classifiers better than expensive global illumination renders, approaching real-data accuracy.

desk verdict A useful NORB-with-3D-scans dataset and a legitimate GAN-selection scheme, but the headline claim of beating global illumination is unsupported because the ensemble comparison is confounded by a 10x larger training set. read the letter →

arxiv 1909.01960 v1 pith:RJGHE5CS submitted 2019-09-04 cs.CV cs.LGstat.ML

Kristofer Schlachter , Connor DeFanti , Sebastian Herscher , Ken Perlin , Jonathan Tompson This is my paper · ORCID

classification cs.CVcs.LGstat.ML

keywords syntheticdatadomainadaptationgenerativeadversarialnetworksglobalilluminationg-bufferrenderinglearnedshadingimageclassificationensemblelearning

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 asks what makes synthetic images useful for training classifiers, and whether physically accurate rendering is necessary. It builds a dataset with real photographs and matching 3D-scanned geometry so that rendering methods can be compared against a real-data baseline directly. The central claim is that an ensemble of generative adversarial network (GAN) refiners, conditioned on geometry buffers and cheap low-sample renders, produces training images that beat full global illumination rendering in downstream classifier accuracy and come close to real data. The lesson is that photorealism is not the right target: learned shading optimized for classifier utility can outperform physically accurate images.

What carries the argument

The central object is the g-buffer, a screen-space encoding of geometry (normals, depth, albedo) that freezes camera pose and object shape while leaving shading free. Conditioning a generative model on this buffer plus a low-sample Monte Carlo image lets the network learn a shading and denoising function without altering semantic content. The ensemble of ten saved GAN refiners, each selected by ranking with a clean classifier trained on real data, expands the dataset tenfold and acts as a regularizer for the downstream classifier.

What would settle it

Train the full global illumination renderer on the same number of images the GAN ensemble uses, for instance by rendering ten independent noisy versions of each scene or ensembling ten GI renders, and compare downstream classifier accuracy; if equally sized GI training sets match or exceed the ensemble, the attribution of the gain to learned shading is wrong.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a learned, partially differentiable shading pipeline gives better domain adaptation than the most physically realistic renderer. Using a fixed g-buffer (screen-space normals, depth, and albedo) to freeze geometry, the authors compare albedo-only rendering, learned spherical-harmonic shading, ambient occlusion, two-bounce direct lighting, and full Monte Carlo global illumination. They then refine low-sample renders with a denoising autoencoder and with GANs. A single GAN refiner roughly matches high-sample GI, while an ensemble of ten refiners reaches 87.33% accuracy on the scanned-object dataset versus 95.01% for real data and 74.74% for high-sample GI; on the RGB validation dataset the ensemble reaches 86.65% accuracy, above the 1024-sample GI baseline. The authors conclude that the best GAN outputs are not the most photorealistic, so conventional photorealism is not a reliable proxy for classifier domain-transfer performance.

Load-bearing premise

The load-bearing premise is that the accuracy gain of the GAN ensemble comes from the learned refinement itself, but the ensemble trains on ten times more images than the GI baseline, which the paper admits is not quite comparable; if the gain is only a data-volume effect, the central claim collapses.

Editorial extensions

If this is right

Cheap low-sample renders plus learned refinement can replace expensive full global illumination as training-data generators for classifiers.
Photorealism is not the correct objective for synthetic training data; images tuned for classifier discriminability can beat physically accurate renders.
Shadows are a key signal for domain transfer: adding shadows improves accuracy far more than adding diffuse shading.
Ensemble expansion of synthetic datasets can improve classifier generalization and even surpass the renderer used as the gold standard.

Reading between the lines

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

The headline comparison is not size-matched: the ensemble trains on ten times more images than the GI baseline, so part or all of the gain could be a data-volume effect rather than a property of learned shading.
The selection protocol keeps the best of three classifier runs for each GAN, which can inflate reported accuracies; a fixed evaluation protocol might show smaller gaps.
The g-buffer conditioning principle suggests a general recipe: any rendering approximation that preserves geometric invariants while leaving shading learnable could be adapted to other downstream tasks such as segmentation or depth estimation.

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 thing to know: the dataset contribution is real and the GAN ensemble idea is plausible, but the central quantitative claim—that the ensemble outperforms full GI rendering—does not hold up as run. The ensemble condition trains on ten times more images than every baseline, and Section 5.1.1 says so explicitly. That is a textbook confound.

What's new: a scanned-3D NORB variant with pixel-aligned g-buffers and one-to-one synthetic/real correspondence, which is genuinely useful for studying domain adaptation and rendering. The GAN selection procedure is also legitimate: they rank refiners by accuracy of a clean classifier trained on real train/val data, never test data. That is not circular, and it improves on the visual-inspection early stopping in Shrivastava et al.

Where it falls short: the headline "surpass full global illumination" comes only from the ensemble rows (NORB 87.33% vs 74.74%; ShapeNet 86.65% vs 82.30%). But the ensemble uses D×10 images. The single-GAN rows sit at parity with the high-sample GI baseline (NORB 75.26% vs 74.74%; ShapeNet 83.59% vs 83.99%). A careful reader should conclude the gain is largely a data-volume effect, not evidence that learned shading beats physically based rendering. Credit where due: the authors admit the comparison is "not quite comparable," yet they still let the headline stand.

Other soft spots: they ran each single-GAN classifier three times and kept the best, with no error bars anywhere. On 4k-10k images, that alone can inflate numbers by a point or two. The ShapeNet "baseline" is itself the high-sample synthetic render, not real data, so "103.17% of baseline" is less impressive than it sounds. And the dataset and code are promised but not released, limiting reproducibility right now.

This paper deserves a serious referee. The dataset concept and the ensemble-selection protocol are worthwhile, and the empirical pattern (single-GAN parity, ensemble gain) is interesting even if the interpretation is not yet supported. I would send it to review with a request for a controlled baseline trained on the same expanded data volume and for error bars, rather than reject it.

Referee Report

2 major / 6 minor

Summary. The paper studies how different synthetic image synthesis techniques affect downstream classifier accuracy as a proxy for domain adaptation. The authors construct a re-rendered version of the NORB benchmark with 3D-scanned geometry, and a ShapeNet-based RGB benchmark, allowing them to compare real images, albedo-only renders, learned spherical-harmonic shading, ambient occlusion, two-bounce direct lighting, Mitsuba global illumination at several sample counts, a denoising autoencoder, and GAN-based refinement. The central claim is that an ensemble of GAN refiners produces training data that outperforms full global illumination rendering and approaches real-data performance, while a single GAN yields near-parity with GI. This claim rests on Tables 2 and 4, where the ensemble column shows large gains (NORB: 87.33% vs. 74.74% for high-sample GI; ShapeNet: 86.65% vs. 82.30% for medium-sample GI).

Significance. If the ensemble claim were validated under controlled conditions, the paper would be significant: it would show that cheap learned refinement can exceed expensive physical rendering for downstream classification, and that conventional photorealism is not the right objective for synthetic training data. The constructed NORB-derived dataset with one-to-one real/synthetic geometry correspondence is a valuable resource for future domain-adaptation research. The use of a clean classifier trained only on real training and validation data to rank GAN models is a sensible and principled model-selection protocol, and the paper is honest about the main confound in Section 5.1.1. However, the current evidence does not yet support the headline causal claim, because the ensemble comparison is not controlled for dataset size.

major comments (2)

[Section 5.1.1, Tables 2 and 4] The ensemble columns in Tables 2 and 4 are not comparable to the Mitsuba, Denoised, or single-GAN rows because the ensemble trains on D×N images with N=10 refiners, while every other row uses D images; the text explicitly states in Section 5.1.1 that 'more training data is used and is therefore not quite comparable.' The headline accuracy gains (NORB: 87.33% vs. 74.74% for high-sample GI; ShapeNet: 86.65% vs. 82.30% for medium-sample GI, 103.17% of baseline) therefore vary dataset size and learned refinement simultaneously. The controlled single-GAN comparisons show only parity or worse (NORB: 75.26% vs. 74.74%; ShapeNet: 83.59% vs. 83.99% absolute baseline), which undercuts the claim that learned shading itself outperforms GI. A control that trains on D unique high-sample GI images, or on D×N augmented or duplicated GI images, is needed to separate the data-volume effect from the refinement effect.
[Section 5.1, Section 6] For every single-GAN evaluation, the paper reports that the classifier was trained three times and the top-performing model was kept, and no variance or statistical test is reported anywhere in Sections 5 or 6. Because the decisive differences are small (for example, 75.26% vs. 74.74% on NORB, and 83.59% vs. 83.99% on ShapeNet), the reported numbers are best-case estimates and may overstate the effect. Report mean and standard deviation over classifier seeds, and ideally over GAN training runs, or provide a paired comparison, before making claims that learned refinement 'surpasses' global illumination.

minor comments (6)

[Section 6.1, Table 1] The text reports 66.89% for the two-bounce rendering condition, while Table 1 lists 66.49%; please reconcile this inconsistency.
[Section 4.2] 'CFAR-10' is a typo for 'CIFAR-10'.
[Section 3] 'Shied et al.' should be 'Schied et al.' in the discussion of spatiotemporal variance-guided filtering.
[Section 6.1] 'This result demonstrations' should be 'This result demonstrates'.
[Figure 11 caption] The caption orders panels as 'Right: ... Middle: ... Left: ...' while the figure text appears to describe the input on the right; please clarify the panel ordering so the reader can follow the comparison.
[Tables 2 and 4] The 'Baseline %' column is used differently in Tables 2 and 4 (as an accuracy value versus a percentage of the baseline), which is confusing; label the columns consistently and state the absolute baseline accuracy in each caption.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: GAN selection uses a clean classifier on real training/validation data, and the ensemble comparison's data-size confound is acknowledged in the paper rather than being a definitional reduction.

full rationale

The paper's central claims are that a GAN ensemble produces synthetic training data that outperforms full GI renders and approaches real-data performance, and that learned shading/denoising can approximate or exceed offline rendering. Checking the derivation chain: the GAN ranking uses a 'clean classifier' trained only on real training and validation data, with test data explicitly excluded; this is a validation-based model-selection procedure, not a prediction of the final test accuracy from fitted values. The final classifiers are retrained independently on synthetic data and evaluated on real test images, so no equation in the paper equates the reported accuracy to a fitted parameter or to the selection criterion. The learned SH coefficients are fit to approximate real image statistics, but the SH experiment is a baseline comparison, not the paper's headline claim. The ensemble uses D*N images while other rows use D images, and the paper itself states 'more training data is used and is therefore not quite comparable'; this is a legitimate experimental confound, but it is not circularity because the reported accuracy is not defined or constructed to equal the dataset-size expansion. There are no load-bearing self-citations: the references to [24], [2], and [3] are external prior works used as architectural inspiration, not as justification of the paper's conclusions. No uniqueness theorem is imported from the authors' own prior work, and no known result is merely renamed. The best-of-three classifier selection for single-GAN rows is a selection-bias concern, not a circular reduction. Accordingly, no circular step can be exhibited, and the derivation is self-contained with respect to the claims it makes.

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

The paper introduces no new physical entities. Its method depends on learned parameters (SH coefficients, denoiser weights, GAN weights) and hand-chosen experimental constants (ensemble size, sample counts, top-10 cutoff). The g-buffer is a standard computer graphics construct, not an invented entity.

free parameters (4)

Spherical harmonic lighting coefficients = Not reported (9 coefficients learned)
The learned SH module is fit to approximate real image statistics; the coefficients are data-driven, not derived from first principles.
Ensemble size N = 10
The number of GAN refiners in the ensemble is chosen by hand with no ablation or sensitivity analysis.
GAN selection cutoff = Top 10 of saved checkpoints
The number of ranked refiners to include is arbitrary and not varied.
Sample counts = NORB: 128/4/1; ShapeNet: 1024/32/10
Chosen as high/medium/low sample counts without systematic justification or error analysis.

assumptions (4)

domain assumption G-buffers (normals, depth, albedo) provide sufficient conditioning information for learned shading and refinement.
Section 4 assumes pixel-aligned geometry is enough to recreate scenes and drive the networks.
domain assumption Classifier accuracy on NORB and ShapeNet is a valid proxy for domain adaptation quality.
This proxy is used throughout to measure the utility of each rendering method.
domain assumption The clean classifier used to rank GANs does not leak test data information.
Section 5.1 states only training and validation data are used, but the best-of-three classifier selection introduces a form of selection bias that is not accounted for.
domain assumption NORB toys have approximately homogeneous matte BRDF, enabling close recreation of real photographs.
Section 4.1 states the paint is uniform matte; this underpins the claimed 1:1 correspondence between synthetic and real domains.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Beyond Photo Realism for Domain Adaptation from Synthetic Data." pith.science (2026). https://pith.science/paper/RJGHE5CS

@misc{pith2026190901960,
  author       = {Pith},
  title        = {Pith review of: Beyond Photo Realism for Domain Adaptation from Synthetic Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RJGHE5CS}},
  note         = {Machine review of arXiv:1909.01960}
}

read the original abstract

As synthetic imagery is used more frequently in training deep models, it is important to understand how different synthesis techniques impact the performance of such models. In this work, we perform a thorough evaluation of the effectiveness of several different synthesis techniques and their impact on the complexity of classifier domain adaptation to the "real" underlying data distribution that they seek to replicate. In addition, we propose a novel learned synthesis technique to better train classifier models than state-of-the-art offline graphical methods, while using significantly less computational resources. We accomplish this by learning a generative model to perform shading of synthetic geometry conditioned on a "g-buffer" representation of the scene to render, as well as a low sample Monte Carlo rendered image. The major contributions are (i) a dataset that allows comparison of real and synthetic versions of the same scene, (ii) an augmented data representation that boosts the stability of learning and improves the datasets accuracy, (iii) three different partially differentiable rendering techniques where lighting, denoising and shading are learned, and (iv) we improve a state of the art generative adversarial network (GAN) approach by using an ensemble of trained models to generate datasets that approach the performance of training on real data and surpass the performance of the full global illumination rendering.

Figures

Figures reproduced from arXiv: 1909.01960 by the authors.

**Figure 2.** Visualization of first 3 bands of the spherical har [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 4.** A sample of the ShapeNet train dataset. The ShapeNet dataset consists of tens of thousands of models. We use a subset of ShapeNet: 10 different models from 10 categories each. The dataset size was chosen to match similar popular, small datasets CFAR-10 [11] and MNIST [12] and since we want to measure the relative performance of our domain adaptation techniques in the regime of limited training data (which more clo… view at source ↗

Figures from the paper (7 more)

**Figure 6.** Figure 6: Spherical Harmonics network architecture. [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]

**Figure 7.** Figure 7: Left: Learned SH. Middle: SH x Albedo. Right: [PITH_FULL_IMAGE:figures/full_fig_p005_7.png]

**Figure 5.** Figure 5: Architecture of VGG08 Batch Norm Classifier [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]

**Figure 8.** Figure 8: Two Bounce Rendering. Note that it is the equiv [PITH_FULL_IMAGE:figures/full_fig_p006_8.png]

**Figure 9.** Figure 9: Mitsuba renders at different sample counts. Left: [PITH_FULL_IMAGE:figures/full_fig_p006_9.png]

**Figure 10.** Figure 10: Architecture of Denoising Model Given that the low sample GI renders take far less time than the high sample renders, we next experiment to see if the lower sample count images can be denoised using a simplified version of the denoising auto-encoder with skip connec…

**Figure 11.** Figure 11: Right: 4 sample Mitsuba image passed into [PITH_FULL_IMAGE:figures/full_fig_p007_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 18 canonical work pages

[1]

Kernel-predicting convolutional networks for denoising monte carlo ren- derings

Steve Bako, Thijs V ogels, Brian McWilliams, Mark Meyer, Jan Nov ´ak, Alex Harvill, Pradeep Sen, Tony DeRose, and Fabrice Rousselle. Kernel-predicting convolutional networks for denoising monte carlo ren- derings. ACM Transactions on Graphics (TOG) (Pro- ceedings of SIGGRAPH 2017), 36(4), July 2017. 2

work page 2017
[2]

Unsuper- vised pixel-level domain adaptation with generative adversarial networks

Konstantinos Bousmalis, Nathan Silberman, David Dohan, Dumitru Erhan, and Dilip Krishnan. Unsuper- vised pixel-level domain adaptation with generative adversarial networks. CoRR, abs/1612.05424, 2016. 2, 6, 7

arXiv 2016
[3]

Alla Chaitanya, Anton S

Chakravarty R. Alla Chaitanya, Anton S. Kaplanyan, Christoph Schied, Marco Salvi, Aaron Lefohn, Derek Nowrouzezahrai, and Timo Aila. Interactive recon- struction of monte carlo image sequences using a re- current denoising autoencoder. ACM Trans. Graph., 36(4):98:1–98:12, July 2017. 2, 6

work page 2017
[4]

Chang, Thomas A

Angel X. Chang, Thomas A. Funkhouser, Leonidas J. Guibas, Pat Hanrahan, Qi-Xing Huang, Zimo Li, Sil- vio Savarese, Manolis Savva, Shuran Song, Hao Su, Jianxiong Xiao, Li Yi, and Fisher Yu. Shapenet: An information-rich 3d model repository. CoRR, abs/1512.03012, 2015. 4

arXiv 2015
[5]

Imagenet: A large-scale hierarchi- cal image database

Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchi- cal image database. In Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on, pages 248–255. IEEE, 2009. 1

work page 2009
[6]

Ganin and V

Y . Ganin and V . Lempitsky. Unsupervised Do- main Adaptation by Backpropagation. ArXiv e-prints, September 2014. 1, 6

work page 2014
[7]

Generative adversarial nets

Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. In Z. Ghahramani, M. Welling, C. Cortes, N. D. Lawrence, and K. Q. Weinberger, editors, Advances in Neural Information Processing Systems 27 , pages 2672–2680. Curran Associates, Inc., 2014. 2

work page 2014
[8]

Hp 3d scan: Hp ofﬁcial website

HP. Hp 3d scan: Hp ofﬁcial website. www8.hp.com/us/en/campaign/3Dscanner/overview.html,

work page

Show all 28 references

[9]

Mitsuba renderer, 2010

Wenzel Jakob. Mitsuba renderer, 2010. http://www.mitsuba-renderer.org. 1

2010
[10]

A machine learning approach for ﬁltering monte carlo noise

Nima Khademi Kalantari, Steve Bako, and Pradeep Sen. A machine learning approach for ﬁltering monte carlo noise. ACM Trans. Graph., 34(4):122:1–122:12, July 2015. 2

2015
[11]

Learning multiple layers of features from tiny images

Alex Krizhevsky. Learning multiple layers of features from tiny images. Technical report, 2009. 4

2009
[12]

Lecun, L

Y . Lecun, L. Bottou, Y . Bengio, and P. Haffner. Gradient-based learning applied to document recog- nition. Proceedings of the IEEE , 86(11):2278–2324, Nov 1998. 4

1998
[13]

Learn- ing methods for generic object recognition with invari- ance to pose and lighting

Yann LeCun, Fu Jie Huang, and L ´eon Bottou. Learn- ing methods for generic object recognition with invari- ance to pose and lighting. In Proceedings of the 2004 IEEE Computer Society Conference on Computer Vi- sion and Pattern Recognition , CVPR’04, pages 97– 104, Washington,...

2004
[14]

Belongie, Lubomir D

Tsung-Yi Lin, Michael Maire, Serge J. Belongie, Lubomir D. Bourdev, Ross B. Girshick, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll ´ar, and C. Lawrence Zitnick. Microsoft COCO: common ob- jects in context. CoRR, abs/1405.0312, 2014. 1

2014 arXiv
[15]

Efﬁcient algorithms for local and global accessibility shading

Gavin Miller. Efﬁcient algorithms for local and global accessibility shading. In Proceedings of the 21st An- nual Conference on Computer Graphics and Inter- active Techniques, SIGGRAPH ’94, pages 319–326, New York, NY , USA, 1994. ACM. 3

1994
[16]

Deep shading: Convolutional neural networks for screen- space shading

Oliver Nalbach, Elena Arabadzhiyska, Dushyant Mehta, Hans-Peter Seidel, and Tobias Ritschel. Deep shading: Convolutional neural networks for screen- space shading. CoRR, abs/1603.06078, 2016. 2

2016 arXiv
[17]

Physically Based Rendering, Second Edition: From Theory To Imple- mentation

Matt Pharr and Greg Humphreys. Physically Based Rendering, Second Edition: From Theory To Imple- mentation. Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, 2nd edition, 2010. 5

2010
[18]

An efﬁ- cient representation for irradiance environment maps

Ravi Ramamoorthi and Pat Hanrahan. An efﬁ- cient representation for irradiance environment maps. In Proceedings of the 28th Annual Conference on Computer Graphics and Interactive Techniques, SIG- GRAPH ’01, pages 497–500, New York, NY , USA,
[19]

Richter, Vibhav Vineet, Stefan Roth, and Vladlen Koltun

Stephan R. Richter, Vibhav Vineet, Stefan Roth, and Vladlen Koltun. Playing for data: Ground truth from computer games. CoRR, abs/1608.02192, 2016. 1, 2

2016 arXiv
[20]

Compre- hensible rendering of 3-d shapes

Takafumi Saito and Tokiichiro Takahashi. Compre- hensible rendering of 3-d shapes. In Proceedings of the 17th Annual Conference on Computer Graphics and Interactive Techniques , SIGGRAPH ’90, pages 197–206, New York, NY , USA, 1990. ACM. 2

1990
[21]

Alla Chaitanya, John Burgess, Shiqiu Liu, Carsten Dachsbacher, Aaron Lefohn, and Marco Salvi

Christoph Schied, Anton Kaplanyan, Chris Wyman, Anjul Patney, Chakravarty R. Alla Chaitanya, John Burgess, Shiqiu Liu, Carsten Dachsbacher, Aaron Lefohn, and Marco Salvi. Spatiotemporal variance- guided ﬁltering: Real-time reconstruction for path- traced global illumination. I...

2017
[22]

Little, and Mark Schmidt

Alireza Shafaei, James J. Little, and Mark Schmidt. Play and learn: Using video games to train computer vision models. CoRR, abs/1608.01745, 2016. 1, 2

2016 arXiv
[23]

Kessenich, and Bill M

Dave Shreiner, Graham Sellers, John M. Kessenich, and Bill M. Licea-Kane. OpenGL Programming Guide: The Ofﬁcial Guide to Learning OpenGL, Ver- sion 4.3. Addison-Wesley Professional, 8th edition,
[24]

Learning from simulated and unsupervised images through ad- versarial training

Ashish Shrivastava, Tomas Pﬁster, Oncel Tuzel, Josh Susskind, Wenda Wang, and Russ Webb. Learning from simulated and unsupervised images through ad- versarial training. CoRR, abs/1612.07828, 2016. 1, 2, 6

2016 arXiv
[25]

Very deep convolutional networks for large-scale image recogni- tion

Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recogni- tion. CoRR, abs/1409.1556, 2014. 5

2014 arXiv
[26]

Ren- dergan: Generating realistic labeled data

Leon Sixt, Benjamin Wild, and Tim Landgraf. Ren- dergan: Generating realistic labeled data. CoRR, abs/1611.01331, 2016. 1, 2

2016 arXiv
[27]

Bovik, Hamid R

Zhou Wang, Alan C. Bovik, Hamid R. Sheikh, and Eero P. Simoncelli. Image quality assessment: From error visibility to structural similarity. IEEE TRANS- ACTIONS ON IMAGE PROCESSING , 13(4):600– 612, 2004. 6

2004
[28]

Funkhouser

Yinda Zhang, Shuran Song, Ersin Yumer, Manolis Savva, Joon-Young Lee, Hailin Jin, and Thomas A. Funkhouser. Physically-based rendering for indoor scene understanding using convolutional neural net- works. CoRR, abs/1612.07429, 2016. 1, 2

2016 arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Kernel-predicting convolutional networks for denoising monte carlo ren- derings

Steve Bako, Thijs V ogels, Brian McWilliams, Mark Meyer, Jan Nov ´ak, Alex Harvill, Pradeep Sen, Tony DeRose, and Fabrice Rousselle. Kernel-predicting convolutional networks for denoising monte carlo ren- derings. ACM Transactions on Graphics (TOG) (Pro- ceedings of SIGGRAPH 2017), 36(4), July 2017. 2

work page 2017

[2] [2]

Unsuper- vised pixel-level domain adaptation with generative adversarial networks

Konstantinos Bousmalis, Nathan Silberman, David Dohan, Dumitru Erhan, and Dilip Krishnan. Unsuper- vised pixel-level domain adaptation with generative adversarial networks. CoRR, abs/1612.05424, 2016. 2, 6, 7

arXiv 2016

[3] [3]

Alla Chaitanya, Anton S

Chakravarty R. Alla Chaitanya, Anton S. Kaplanyan, Christoph Schied, Marco Salvi, Aaron Lefohn, Derek Nowrouzezahrai, and Timo Aila. Interactive recon- struction of monte carlo image sequences using a re- current denoising autoencoder. ACM Trans. Graph., 36(4):98:1–98:12, July 2017. 2, 6

work page 2017

[4] [4]

Chang, Thomas A

Angel X. Chang, Thomas A. Funkhouser, Leonidas J. Guibas, Pat Hanrahan, Qi-Xing Huang, Zimo Li, Sil- vio Savarese, Manolis Savva, Shuran Song, Hao Su, Jianxiong Xiao, Li Yi, and Fisher Yu. Shapenet: An information-rich 3d model repository. CoRR, abs/1512.03012, 2015. 4

arXiv 2015

[5] [5]

Imagenet: A large-scale hierarchi- cal image database

Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchi- cal image database. In Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on, pages 248–255. IEEE, 2009. 1

work page 2009

[6] [6]

Ganin and V

Y . Ganin and V . Lempitsky. Unsupervised Do- main Adaptation by Backpropagation. ArXiv e-prints, September 2014. 1, 6

work page 2014

[7] [7]

Generative adversarial nets

Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. In Z. Ghahramani, M. Welling, C. Cortes, N. D. Lawrence, and K. Q. Weinberger, editors, Advances in Neural Information Processing Systems 27 , pages 2672–2680. Curran Associates, Inc., 2014. 2

work page 2014

[8] [8]

Hp 3d scan: Hp ofﬁcial website

HP. Hp 3d scan: Hp ofﬁcial website. www8.hp.com/us/en/campaign/3Dscanner/overview.html,

work page