Pith. sign in

REVIEW 4 major objections 5 minor 46 references

DTSGAN: Learning Dynamic Textures via Spatiotemporal Generative Adversarial Network

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A single-video 3D GAN with a sequential data-update strategy synthesizes diverse dynamic textures and outperforms three prior methods on quality and smoothness.

desk verdict A straightforward video extension of SinGAN with a sensible training trick, but the quantitative evidence for its main claim is not trustworthy. read the letter →

arxiv 2412.16948 v1 pith:3IUKMGOZ submitted 2024-12-22 cs.CV

Xiangtian Li , Xiaobo Wang , Zhen Qi , Han Cao , Zhaoyang Zhang , Ao Xiang This is my paper · ORCID
classification cs.CV
keywords dynamictexturesynthesisspatiotemporalGANsingle-videolearningdataupdatestrategymulti-scalepyramid3Dconvolutionvideogenerationdiversity
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 introduces DTSGAN, a generative adversarial network that learns to synthesize dynamic textures from a single video clip by building a coarse-to-fine pyramid of 3D generators and discriminators. The authors claim that this spatiotemporal architecture captures both content and motion, and that a sequential data-update strategy—sliding the training window along the video—improves the diversity of generated clips without causing mode collapse. On 18 videos from the Dynamic Texture Database, they report the highest MS-SSIM and lowest FID and δ-N-LPIPS among Two-Stream, DG, and cgCNN, and a user study in which 53.89% of preferences favored DTSGAN. The stated limitation is that a model must be trained from scratch for each texture video.

What carries the argument

The central object is a multiscale pyramid of 3D GANs, where the coarsest generator maps Gaussian noise to a video and each finer generator receives the upsampled output of the previous scale plus fresh noise, so that new details are added progressively. At scale $n$, the generator computes $\tilde{x}_n = G_n(z_n, (\tilde{x}_{n-1})^{\uparrow r})$ for $n > 0$, and the coarsest scale computes $\tilde{x}_0 = G_0(z_0)$. The second load-bearing mechanism is the data update strategy: instead of retraining on one fixed clip or randomly switching clips, the training window shifts to its neighbor every fixed number of steps, exposing the model to more motions while keeping the content distribution stable.

What would settle it

Train a modern video GAN such as MoCoGAN or DVD-GAN on the same 18 DTDB videos and compute the same metrics; if the reported margins disappear or reverse, the claim of favorable performance fails.

Watch

Extended reading notes

Core claim

On its own terms, DTSGAN can generate 16-frame, 150×150 dynamic texture videos from a single training clip, and its sequential neighbor-window data update strategy yields more diverse outputs than training on one fixed clip alone. Compared with Two-Stream, DG, and cgCNN, the model reportedly achieves the best average MS-SSIM (0.621), best FID (193.063), and best δ-N-LPIPS (0.223) on the 18 selected DTDB videos, while the user study records 53.89% preference for DTSGAN. The architecture extends a single-image pyramid generator to 3D by using 3D convolutions and a WGAN-GP adversarial loss, with reconstruction loss at each scale to stabilize training.

Load-bearing premise

The evaluation assumes that Two-Stream, DG, and cgCNN are strong, fairly tuned baselines and that frame-wise metrics (FID, MS-SSIM averaged per frame, and a smoothness measure) capture dynamic texture quality.

Editorial extensions

If this is right

  • If the reported margins hold, DTSGAN advances single-video dynamic texture synthesis by combining multi-scale spatial modeling with temporal data updates.
  • The sequential neighbor-window update strategy is a direct, testable recipe for improving diversity in other single-example generators.
  • The encoder-based future prediction extension shows that the learned spatiotemporal representation can condition generation on a starting frame.
  • The copy-paste editing experiment indicates the model can fuse reference patches into coherent dynamic textures.
  • Training from scratch per video remains a practical bottleneck, limiting direct deployment to new textures.

Reading between the lines

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

  • The diversity gains may come less from the specific neighbor rule than from simply training on more distinct clips; a randomized-window control would isolate the mechanism.
  • Because the reported metrics average frame-level FID and MS-SSIM, they may not capture temporal coherence as strongly as the δ-N-LPIPS smoothness measure; a direct optical-flow consistency test would be a sharper comparison.
  • The data-update strategy effectively performs curriculum learning over time; the same idea could be applied to other single-video tasks such as video extrapolation or unconditional video generation from long recordings.
  • If the method is robust to the choice of update frequency (50–200 epochs), it could be adapted to unsupervised video summarization or anomaly detection by learning the texture distribution of normal footage.
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 / 5 minor

Summary. The paper proposes DTSGAN, a multi-scale 3D generative adversarial network trained on a single 16-frame dynamic texture video. The model extends SinGAN to the temporal domain, building a pyramid of generators and discriminators from coarse to fine scales, and introduces a sequential data update strategy that alternates among neighboring clips during training to improve output diversity. The authors report qualitative and quantitative comparisons with Two-Stream, DG, and cgCNN on selected DTDB videos, claiming superior MS-SSIM, FID, and δ-N-LPIPS scores, and a user study in which DTSGAN is preferred. The paper also lists future prediction and video editing as applications, though without experimental support.

Significance. The paper addresses an interesting and still active problem, and the proposed architecture is a plausible extension of SinGAN to video: a pyramid of 3D GANs trained on a single clip, with a sequential data update strategy to increase diversity. If the method worked as claimed, it could be a useful contribution to dynamic texture synthesis. The ablation in Table 3 provides some evidence that the data update strategy improves LPIPS-based diversity on the five categories tested. However, the central comparative claim rests on an evaluation that is not statistically or metrologically sound, and the paper omits any evaluation of the two additional applications it lists. The method description is also incomplete, with a garbled adversarial loss equation and a missing metric definition.

major comments (4)
  1. [Sec. 4.3, Table 1] The quantitative evaluation in Table 1 does not support the claim that DTSGAN outperforms the baselines. MS-SSIM is averaged per corresponding frame between the generated clip and the training exemplar, so it rewards pixel-level copying and penalizes diverse but plausible samples. FID is computed between only 16 generated frames and 16 reference frames, which is too few for a reliable Inception-space covariance estimate. δ-N-LPIPS, as defined, rewards small frame-to-frame LPIPS differences and can be minimized by a static clip, so it does not measure 'natural motion.' No standard deviations, per-video breakdowns, or significance tests are reported, and no direct temporal metric such as FVD is used. A static-frame control baseline should be included.
  2. [Sec. 4.1, baselines] The baseline set is not representative of the current state of the art. The compared methods are Two-Stream (2018), DG (2019), and cgCNN (2021), while MoCoGAN and DVD-GAN are cited in Section 2 but are not included in the experiments. The contribution 'performs favorably against the state-of-the-art methods' is therefore not established by the reported comparisons.
  3. [Sec. 3.2-3.4 and Eq. (6)] The method and metric definitions are incomplete or garbled. Equation (3), intended to be the WGAN-GP adversarial loss, is rendered as 'LDn =Dn(xn)−Dn(Gn(x˜n nDn(xˆn)∥2 −1)2,' with missing expectation terms and an unclear gradient penalty. Equation (6), which defines δ-N-LPIPS, is absent from the text though referenced in Section 4.3. These omissions make the training objective and a core reported metric irreproducible.
  4. [Sec. 1 and Sec. 5] Future prediction and video editing are listed as contributions and illustrated in Figure 1, but the manuscript presents no experiments, ablations, or quantitative results for either application. Claims about these applications are therefore unsupported.
minor comments (5)
  1. [Sec. 4.1, scales] The scale parameters are inconsistent: with a coarsest scale of 25px, a finest scale of 150px, and 8 scales, the scaling factor should be approximately 1.25 rather than r ≈ 1.39; please clarify the intended relationship.
  2. [References] The reference list contains numerous citations that do not correspond to the cited context, including [3], [10]-[12], [14], [16]-[17], [19], [23], [26]-[28], [30], [32], [37], and [44]-[46]; the bibliography should be thoroughly cleaned.
  3. [Whole manuscript] The manuscript has many typographical and formatting errors, such as missing spaces ('Synthesizingadynamictextureisanintricateprocessinspaceandtime.') and corrupted notation ('x˜r n ec' in Eq. (4)); a careful proofreading pass is needed.
  4. [Sec. 4.1, data update frequency] The data update frequency is reported only as '50-200 is a fit range of epochs' with no sensitivity analysis or selection criterion, which makes this key hyperparameter difficult to reproduce.
  5. [Table 2, user study] The user study reports only aggregate preference percentages without confidence intervals or significance tests, so the statistical strength of the preference for DTSGAN is unclear.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: DTSGAN is an empirical single-video GAN pipeline whose central claims rest on external baselines and standard evaluation metrics, not on self-referential reasoning.

full rationale

The paper makes no formal derivation whose conclusion is equivalent to its inputs. Section 3 defines the multi-scale generators by Eq. (1) and (2), trains them with a WGAN-GP adversarial loss (Eq. 3) and a reconstruction loss (Eq. 4), and introduces a sequential data-update strategy for diversity. The central claim in Section 4.3 and Table 1 is empirical: DTSGAN has the highest MS-SSIM and lowest FID and delta-N-LPIPS among the tested methods. No hyperparameter or parameter is fitted to force these metric values; MS-SSIM, FID, and delta-N-LPIPS are computed post hoc and are not part of the training loss. The data-update strategy is evaluated with LPIPS diversity in Table 3, which is a standard external measure of the property the strategy targets, not a reduction of the target into the training objective. The architecture borrows from SinGAN, VideoGAN, and WGAN-GP, which are external prior works. Some references include the authors' own unrelated papers, but none is load-bearing for the architecture, the training losses, or the evaluation protocol. The stated limitation in Section 5, namely that the method must be retrained from scratch for every texture video, is an honest caveat and not evidence of circularity. Concerns that MS-SSIM rewards similarity to the training clip, that FID with 16 frames is statistically weak, and that delta-N-LPIPS rewards static clips are validity concerns about the benchmark, not instances of a prediction reducing to its input by construction. Verdict: no significant circularity.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The method introduces no new physical or mathematical entities. It does introduce an empirical training heuristic, the sequential data update, which depends on several hand-set hyperparameters. The abstract claims a single-video model, but the data update strategy actually trains on a sequence of clips from one video. The hidden hyperparameters and unreported loss values are the main free parameters.

free parameters (6)
  • Scaling factor r = ~1.39
    Upsampling ratio between pyramid scales; hand-set to produce 8 scales from 25px to 150px.
  • Number of scales N = 8
    Chosen so that the coarsest scale is 25px and the finest is 150px.
  • Data update frequency = 50-200 epochs
    Reported as a 'fit range' that balances diversity and quality; the choice is empirical and affects results.
  • Reconstruction loss weight eta = not reported
    Weights reconstruction loss in Eq. (5); no value is given, making replication impossible.
  • WGAN-GP penalty coefficient lambda = not reported
    Used in the adversarial loss in Eq. (3); value not specified.
  • Network depth, channels, kernel sizes, learning rate = not reported
    Required to train the model but absent from Section 3.2 and Section 4.1.
assumptions (4)
  • domain assumption A dynamic texture can be represented as a pyramid of scales, each sharing the same patch distribution, extending SinGAN's image assumption to video.
    Section 3.1-3.2; this is the core modeling premise and is adopted from image synthesis without independent validation for video.
  • domain assumption A 3D receptive field of 11x11x11 is sufficient to capture both local appearance and temporal motion in a 16-frame clip.
    Section 3.2-3.3 states the receptive field size but provides no analysis or ablation justifying it.
  • domain assumption Sequentially selecting neighboring video clips from the same source keeps training stable while exposing more motion, unlike random clipping.
    Section 3.3 claims random clipping fails and neighbor clipping works, but offers only one qualitative figure and no formal characterization.
  • domain assumption WGAN-GP provides steady training for this multi-scale 3D architecture.
    Section 3.4 uses WGAN-GP loss from [38] without verification that it behaves well in this setting.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DTSGAN: Learning Dynamic Textures via Spatiotemporal Generative Adversarial Network." pith.science (2026). https://pith.science/paper/3IUKMGOZ

@misc{pith2026241216948,
  author       = {Pith},
  title        = {Pith review of: DTSGAN: Learning Dynamic Textures via Spatiotemporal Generative Adversarial Network},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3IUKMGOZ}},
  note         = {Machine review of arXiv:2412.16948}
}
read the original abstract

Dynamic texture synthesis aims to generate sequences that are visually similar to a reference video texture and exhibit specific stationary properties in time. In this paper, we introduce a spatiotemporal generative adversarial network (DTSGAN) that can learn from a single dynamic texture by capturing its motion and content distribution. With the pipeline of DTSGAN, a new video sequence is generated from the coarsest scale to the finest one. To avoid mode collapse, we propose a novel strategy for data updates that helps improve the diversity of generated results. Qualitative and quantitative experiments show that our model is able to generate high quality dynamic textures and natural motion.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 40 canonical work pages

  1. [1]

    Graphcut textures: Image and video synthesis using graph 9 cuts[J]

    Kwatra V, Schödl A, Essa I, et al. Graphcut textures: Image and video synthesis using graph 9 cuts[J]. Acm transactions on graphics (tog), 2003, 22(3): 277-286

    work page 2003

  2. [2]

    Local spatiotemporal features for dynamic texture synthesis[J]

    Lizarraga-Morales R A, Guo Y, Zhao G, et al. Local spatiotemporal features for dynamic texture synthesis[J]. EURASIP Journal on Image and Video Processing, 2014, 2014: 1-15

    work page 2014

  3. [3]

    An intelligent electronic lock for remote-control system based on the internet of things[C]//journal of physics: conference series

    Li W, Li H, Gong A, et al. An intelligent electronic lock for remote-control system based on the internet of things[C]//journal of physics: conference series. IOP Publishing, 2018, 1069(1): 012134

    work page 2018

  4. [4]

    Very deep convolutional networks for large-scale image recognition[J]

    Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:1409.1556, 2014

    arXiv 2014

  5. [5]

    Yuan L, Wen F, Liu C, et al. Synthesizing dynamic texture with closed-loop linear dynamic system[C]//Computer Vision-ECCV 2004: 8th European Conference on Computer Vision, Prague, Czech Republic, May 11-14, 2004. Proceedings, Part II 8. Springer Berlin Heidelberg, 2004: 603-616

    work page 2004

  6. [6]

    Synthesising Dynamic Textures using Convolutional Neural Networks

    Funke C M, Gatys L A, Ecker A S, et al. Synthesising dynamic textures using convolutional neural networks[J]. arXiv preprint arXiv:1702.07006, 2017

    work page Pith review arXiv 2017

  7. [7]

    Two-stream convolutional networks for dynamic texture synthesis[C]//Proceedings of the IEEE conference on computer vision and pattern recognition

    Tesfaldet M, Brubaker M A, Derpanis K G. Two-stream convolutional networks for dynamic texture synthesis[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 6703-6712

    work page 2018

  8. [8]

    Dynamic texture modeling and synthesis using multi-kernel Gaussian process dynamic model[J]

    Zhu Z, You X, Yu S, et al. Dynamic texture modeling and synthesis using multi-kernel Gaussian process dynamic model[J]. Signal Processing, 2016, 124: 63-71

    work page 2016

Show all 46 references
  1. [9]

    Generating videos with scene dynamics[J]

    Vondrick C, Pirsiavash H, Torralba A. Generating videos with scene dynamics[J]. Advances in neural information processing systems, 2016, 29

    2016

  2. [10]

    Improved YOLOv5 Based on Attention Mechanism and FasterNet for Foreign Object Detection on Railway and Airway tracks[J]

    Qi Z, Ma D, Xu J, et al. Improved YOLOv5 Based on Attention Mechanism and FasterNet for Foreign Object Detection on Railway and Airway tracks[J]. arXiv preprint arXiv:2403.08499, 2024

    2024 arXiv

  3. [11]

    A neural matrix decomposition recommender system model based on the multimodal large language model[J]

    Xiang A, Huang B, Guo X, et al. A neural matrix decomposition recommender system model based on the multimodal large language model[J]. arXiv preprint arXiv:2407.08942, 2024

    2024 arXiv

  4. [12]

    Transformer-Based Classification Outcome Prediction for Multimodal Stroke Treatment[J]

    Ma D, Wang M, Xiang A, et al. Transformer-Based Classification Outcome Prediction for Multimodal Stroke Treatment[J]. arXiv preprint arXiv:2404.12634, 2024

    2024 arXiv

  5. [13]

    Conditional generative ConvNets for exemplar-based texture synthesis[J]

    Wang Z M, Li M H, Xia G S. Conditional generative ConvNets for exemplar-based texture synthesis[J]. IEEE Transactions on Image Processing, 2021, 30: 2461-2475

    2021

  6. [14]

    A Multimodal Fusion Network For Student Emotion Recognition Based on Transformer and Tensor Product[J]

    Xiang A, Qi Z, Wang H, et al. A Multimodal Fusion Network For Student Emotion Recognition Based on Transformer and Tensor Product[J]. arXiv preprint arXiv:2403.08511, 2024

    2024 arXiv

  7. [15]

    Dynamic textures[C]//Proceedings Eighth IEEE International Conference on Computer Vision

    Soatto S, Doretto G, Wu Y N. Dynamic textures[C]//Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001. IEEE, 2001, 2: 439-446

    2001

  8. [16]

    Improved Unet model for brain tumor image segmentation based on ASPP-coordinate attention mechanism[J]

    Wang Z, Chen Y, Wang F, et al. Improved Unet model for brain tumor image segmentation based on ASPP-coordinate attention mechanism[J]. arXiv preprint arXiv:2409.08588, 2024

    2024 arXiv

  9. [17]

    AI-based NLP section discusses the application and effect of bag-of- words models and TF-IDF in NLP tasks[J]

    Dai S, Li K, Luo Z, et al. AI-based NLP section discusses the application and effect of bag-of- words models and TF-IDF in NLP tasks[J]. Journal of Artificial Intelligence General science (JAIGS) ISSN: 3006-4023, 2024, 5(1): 13-21

    2024

  10. [18]

    Learning dynamic generator model by alternating back-propagation through time[C]//Proceedings of the AAAI Conference on Artificial Intelligence

    Xie J, Gao R, Zheng Z, et al. Learning dynamic generator model by alternating back-propagation through time[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2019, 33(01): 5498- 5507

    2019

  11. [19]

    MPGAAN: Effective and Efficient Heterogeneous Information Network Classification[J]

    Wu Z. MPGAAN: Effective and Efficient Heterogeneous Information Network Classification[J]. Journal of Computer Science and Technology Studies, 2024, 6(4): 08-16

    2024

  12. [20]

    Improving text-image matching with adversarial learning and circle loss for multi-modal steganography[C]//International Workshop on Digital Watermarking

    Hu Y, Cao H, Yang Z, et al. Improving text-image matching with adversarial learning and circle loss for multi-modal steganography[C]//International Workshop on Digital Watermarking. Cham: Springer International Publishing, 2020: 41-52

    2020

  13. [21]

    Ingan: Capturing and remapping the" dna" of a natural image[J]

    Shocher A, Bagon S, Isola P, et al. Ingan: Capturing and remapping the" dna" of a natural image[J]. arXiv preprint arXiv:1812.00231, 2018

    2018 arXiv

  14. [22]

    Singan: Learning a generative model from a single natural image[C]//Proceedings of the IEEE/CVF international conference on computer vision

    Shaham T R, Dekel T, Michaeli T. Singan: Learning a generative model from a single natural image[C]//Proceedings of the IEEE/CVF international conference on computer vision. 2019: 4570- 4580

    2019

  15. [23]

    Stock price prediction based on hybrid CNN-LSTM model

    Zhao, Qinghe, Yue Hao, and Xuechen Li. "Stock price prediction based on hybrid CNN-LSTM model." Applied and Computational Engineering 104 (2024): 110-115

    2024

  16. [24]

    Lin J, Pang Y, Xia Y, et al. Tuigan: Learning versatile image-to-image translation with two unpaired images[C]//Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part IV 16. Springer International Publishing, 2020: 18-35

    2020

  17. [25]

    Deep single image manipulation[J]

    Zabari Y V E H N, Hoshen Y. Deep single image manipulation[J]. arXiv preprint arXiv:2007.01289, 2020

    2007 arXiv

  18. [26]

    The application of artificial intelligence technology in assembly techniques within the industrial sector[J]

    Hong B, Zhao P, Liu J, et al. The application of artificial intelligence technology in assembly techniques within the industrial sector[J]. Journal of Artificial Intelligence General science (JAIGS) ISSN: 3006-4023, 2024, 5(1): 1-12

    2024

  19. [27]

    Multi-modal steganography based on semantic 10 relevancy[C]//International Workshop on Digital Watermarking

    Hu Y, Yang Z, Cao H, et al. Multi-modal steganography based on semantic 10 relevancy[C]//International Workshop on Digital Watermarking. Cham: Springer International Publishing, 2020: 3-14

    2020

  20. [28]

    Deep Learning with Improved Metaheuristic Optimization for Traffic Flow Prediction[J]

    Wu Z. Deep Learning with Improved Metaheuristic Optimization for Traffic Flow Prediction[J]. Journal of Computer Science and Technology Studies, 2024, 6(4): 47-53

    2024

  21. [29]

    Visual dynamics: Probabilistic future frame synthesis via cross convolutional networks[J]

    Xue T, Wu J, Bouman K, et al. Visual dynamics: Probabilistic future frame synthesis via cross convolutional networks[J]. Advances in neural information processing systems, 2016, 29

    2016

  22. [30]

    Research on autonomous driving decision-making strategies based deep reinforcement learning[J]

    Wang Z, Yan H, Wei C, et al. Research on autonomous driving decision-making strategies based deep reinforcement learning[J]. arXiv preprint arXiv:2408.03084, 2024

    2024 arXiv

  23. [31]

    Temporal generative adversarial nets with singular value clipping[C]//Proceedings of the IEEE international conference on computer vision

    Saito M, Matsumoto E, Saito S. Temporal generative adversarial nets with singular value clipping[C]//Proceedings of the IEEE international conference on computer vision. 2017: 2830-2839

    2017

  24. [32]

    Gao H, Wang H, Feng Z, et al. A novel texture extraction method for the sedimentary structures’ classification of petroleum imaging logging[C]//Pattern Recognition: 7th Chinese Conference, CCPR 2016, Chengdu, China, November 5-7, 2016, Proceedings, Part II 7. Springer Singapor...

    2016

  25. [33]

    Mocogan: Decomposing motion and content for video generation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition

    Tulyakov S, Liu M Y, Yang X, et al. Mocogan: Decomposing motion and content for video generation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 1526-1535

    2018

  26. [34]

    Large Scale GAN Training for High Fidelity Natural Image Synthesis[J]

    Brock A. Large Scale GAN Training for High Fidelity Natural Image Synthesis[J]. arXiv preprint arXiv:1809.11096, 2018

    2018 arXiv

  27. [35]

    Adversarial video generation on complex datasets[J]

    Clark A, Donahue J, Simonyan K. Adversarial video generation on complex datasets[J]. arXiv preprint arXiv:1907.06571, 2019

    1907 arXiv

  28. [36]

    Batch normalization: Accelerating deep network training by reducing internal covariate shift[J]

    Ioffe S. Batch normalization: Accelerating deep network training by reducing internal covariate shift[J]. arXiv preprint arXiv:1502.03167, 2015

    2015 arXiv

  29. [37]

    Editable neural radiance fields convert 2D to 3D furniture texture[J]

    Tan C, Wang C, Lin Z, et al. Editable neural radiance fields convert 2D to 3D furniture texture[J]. International Journal of Engineering and Management Research, 2024, 14(3): 62-65

    2024

  30. [38]

    Improved training of wasserstein gans[J]

    Gulrajani I, Ahmed F, Arjovsky M, et al. Improved training of wasserstein gans[J]. Advances in neural information processing systems, 2017, 30

    2017

  31. [39]

    A new large scale dynamic texture dataset with application to convnet understanding[C]//Proceedings of the European Conference on Computer Vision (ECCV)

    Hadji I, Wildes R P. A new large scale dynamic texture dataset with application to convnet understanding[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 320- 335

    2018

  32. [40]

    Multiscale structural similarity for image quality assessment[C]//The Thrity-Seventh Asilomar Conference on Signals, Systems & Computers, 2003

    Wang Z, Simoncelli E P, Bovik A C. Multiscale structural similarity for image quality assessment[C]//The Thrity-Seventh Asilomar Conference on Signals, Systems & Computers, 2003. Ieee, 2003, 2: 1398-1402

    2003

  33. [41]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium[J]

    Heusel M, Ramsauer H, Unterthiner T, et al. Gans trained by a two time-scale update rule converge to a local nash equilibrium[J]. Advances in neural information processing systems, 2017, 30

    2017

  34. [42]

    Continuous and diverse image-to-image translation via signed attribute vectors[J]

    Mao Q, Tseng H Y, Lee H Y, et al. Continuous and diverse image-to-image translation via signed attribute vectors[J]. International Journal of Computer Vision, 2022, 130(2): 517-549

    2022

  35. [43]

    The unreasonable effectiveness of deep features as a perceptual metric[C]//Proceedings of the IEEE conference on computer vision and pattern recognition

    Zhang R, Isola P, Efros A A, et al. The unreasonable effectiveness of deep features as a perceptual metric[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 586 -595

    2018

  36. [44]

    The application of augmented reality (ar) in remote work and education[J]

    Li K, Xirui P, Song J, et al. The application of augmented reality (ar) in remote work and education[J]. arXiv preprint arXiv:2404.10579, 2024

    2024 arXiv

  37. [45]

    Utilizing deep learning to optimize software development processes[J]

    Li K, Zhu A, Zhou W, et al. Utilizing deep learning to optimize software development processes[J]. arXiv preprint arXiv:2404.13630, 2024

    2024 arXiv

  38. [46]

    Task allocation planning based on hierarchical task network for national economic mobilization[J]

    Zhao P, Li K, Hong B, et al. Task allocation planning based on hierarchical task network for national economic mobilization[J]. Journal of Artificial Intelligence General science (JAIGS) ISSN: 3006-4023, 2024, 5(1): 22-31

    2024

Pith tools

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