Diese Präsentation wurde erfolgreich gemeldet.
Wir verwenden Ihre LinkedIn Profilangaben und Informationen zu Ihren Aktivitäten, um Anzeigen zu personalisieren und Ihnen relevantere Inhalte anzuzeigen. Sie können Ihre Anzeigeneinstellungen jederzeit ändern.

Generative Adversarial Networksの基礎と応用について

1.385 Aufrufe

Veröffentlicht am

東北大学で行われた研究会で用いたスライドです。参考になれば幸いです。

Veröffentlicht in: Technologie
  • Login to see the comments

Generative Adversarial Networksの基礎と応用について

  1. 1. 

  2. 2. 
 
 !
  3. 3. 
 
 https://medium.com/@crosssceneofwindff
  4. 4. 
 

  5. 5. 
 
 
 
 
 
 

  6. 6. 
 

  7. 7. θ p(x; θ)
  8. 8. 
 z ~ pz
  9. 9. G: Generator
 D: Discriminator
 V: Objective Function
 
 pz: Noise distribution
 pg: Generator distribution
 pdata: Training data distribution
  10. 10. D(x) = 1 D(G(z)) = 0
  11. 11. D(G(z)) = 1
  12. 12. G D F(y) = alog(y) + blog(1-y) D G(z)))dz
  13. 13. Jensen-Shannon Divergence: D JSD(p||q) = {KL(p||m) + KL(q||m)}/2 where m = (p + q)/2
  14. 14. D pdata = pg 

  15. 15. z ~ pz
  16. 16. 
 
 
 CNN
  17. 17.
  18. 18. 
 

  19. 19.
  20. 20. 
 
 
 

  21. 21. z
  22. 22. Mode Collapse

  23. 23. 
 
 
 
 
 
 
 

  24. 24.
  25. 25. 
 
 

  26. 26. 1-α
  27. 27.
  28. 28.
  29. 29. 
 Generator
  30. 30. Conv1_1 Conv1_2 Conv2_1 Conv2_2 Conv3_1 Conv3_2 Conv3_3 Conv3_4 Conv4_1 Conv4_2 Conv4_3 Conv4_4 Conv5_1 Conv5_2 Conv5_3 Conv5_4 Vgg19(fix)
  31. 31. 
 Encoder Classifier x z y
  32. 32. {xi}i = 1...N ∈ X (X: source domain)
 {yj}j = 1...M ∈ Y (Y: target domain) X Y, Y X 
 
 pix2pix CycleGAN
  33. 33. - G: X → Y - F: Y → X - DX : - DY: 
 

  34. 34. G x Dy F(G(x)) ~ x F y Dx G(F(y)) ~ y
  35. 35. 
 
 Illustration2Vec Kawaii Illustration Kawaii Illustration Kawaii Illustration
  36. 36. Kawaii gif Kawaii Illustration
  37. 37. Kawaii Illustration Kawaii gif
  38. 38. 
 

  39. 39. 
 

  40. 40. 
 

  41. 41.
  42. 42. 
 
 
 
 
 

  43. 43. G: X → Y G Y Y 

  44. 44. 
 

  45. 45. Original Results Id-loss Id-loss A B
  46. 46.
  47. 47. 
 Adversarial loss L1 Distance
  48. 48. σ: µ: source target AdaIN
  49. 49. CycleGAN-VC 1. Network Architecture: 1D CNN -> 2D-1D-2D CNN
 
 
 
 
 
 
 
 
 
 2. Two-Step Adversarial loss
 Adversarial Loss 3. PatchGAN
 Discriminator
  50. 50. StarGAN-VC 1. Source-and-Target Adversarial Loss
 
 
 
 
 
 
 2. Conditional Instance Normalization
 Generator Discriminator Target Source Instance Normalization
  51. 51. ! Ian J Goodfellow, et al., “Generative Adversarial Nets”. NIPS2014 ! Alec Radford et al., “Unsupervised Representation Learning with Deep Convolutional Adversarial Networks”. ICLR2016 ! Naveen Kodali, et al., “On Convergence and Stability of GANs”. arXiv:1705.07215 ! Xudong Mao, et al., “Least Squares Generative Adversarial Networks”. ICCV2016 ! Takeru Miyato, et al., “Spectral Normalization for Generative Adversarial Networks”. ICLR2018 ! Martin Heusel, et al., “GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium”. NIPS2017 ! Lars Mescheder, et al., “Which Training Methods for GANs do actually Converge?”. ICML2018 ! Alexia Jolicoeur-martineau. “The Relativistic Discriminator: A key element missing from Standard GAN”. ICLR2019 ! Martin Arjovsky, et al., “Wasserstein GAN”. arXiv: 1701.07875 ! Ishaan Gulrajani, et al., “Improved Training of Wasserstein GANs”. NIPS2017 ! Akash Srivastava, et al., “VEEGAN: Reducing Mode Collapse in GANs using Implicit Variational Learning”, NIPS2017 ! Chang Xiao, et al., “BourGAN: Generative Networks with Metric Embeddings”. NIPS2018 ! Tong Che, et al., “Mode Regularized Generative Adversarial Networks”. ICLR2017 ! Luke Metz, et al., “Unrolled Generative Adversarial Networks”. ICLR2017 ! Qi Mao, et al., “Mode Seeking Generative Adversarial Networks for Diverse Image Synthesis”. CVPR2019 ! Han Zhang, et al., “StackGAN: Text to Photo-realistic Image Synthesis with Stacked Generative Adversarial Networks” ICCV2017 ! Han Zhang, et al., “StackGAN++: Realistic Image Synthesis with Stacked Generative Adversarial Networks”. TPAMI2018
  52. 52. ! Tero Karras, et al., “Progressive Growing of GANs for Improved Quality, Stability, and Variation”. ICLR2018 ! Andrew Brock, et al., “Large Scale GAN Training for High Fidelity Natural Image Synthesis”. ICLR2019 ! Tero Karras, et al., “A Style-Based Generator Architecture for Generative Adversarial Networks”. CVPR2019 ! Christian Ledig, et al., “Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network”. CVPR2017 ! Xintao Wang, et al., “ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks”. ECCV2018 ! Mengyu Chu, et al., “Temporally Coherent GANs for Video Super-Resolution (TecoGAN)”. arXiv: 1811.0939 ! Phillip Isola, et al., “Image-to-Image Translation with Conditional Adversarial Networks”. CVPR2017 ! Jun-Yan Zhu, et al., “Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks”. ICCV2017 ! Yunjey Choi, et al., “StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation”. CVPR2018 ! Ming-Yu Liu, et al., “Few-Shot Unsupervised Image-to-Image Translation”. ICCV2019 ! Sangwoo Mo, et al., “InstaGAN: Instance-aware Image-to-Image Translation”. ICLR2019 ! Junho Kim, et al., “U-GAT-IT: Unsupervised Generative Attentional Networks with Adaptive Layer-Instance Normalization for Image-to-Image Translation”. arXiv: 1907.10830 ! Eric Tzeng, et al., “Adversarial Discriminative Domain Adaptation”. CVPR2017 ! Issam Laradji, et al., “M-ADDA: Unsupervised Domain Adaptation with Deep Metric Learning”. ICML2018
  53. 53. ! Judy Hoffman, et al., “CyCADA: Cycle-Consistent Adversarial Domain Adaptation”. ICML2018 ! Ming-Yu Liu, et al., “Coupled Generative Adversarial Networks”. NIPS2016 ! Carl Vondrick, et al., “Generating Videos with Scene Dynamics”. NIPS2016 ! Masaki Saito, et al., “Temporal Generative Adversarial Nets with Singular Value Clipping”. ICCV2017 ! Sergey Tulyakov, et al., “MoCoGAN: Decomposing Motion and Content for Video Generation”. CVPR2018 ! Katsunori Ohnishi, et al., “Hierarchical Video Generation from Orthogonal Information: Optical Flow and Texture”. AAAI2018 ! Aidan Clark, et al., “Adversarial Video Generation on Complex Datasets”. arXiv: 1907.06571 ! Jiajun Wu, et al., “Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling”. NIPS2016 ! Ruihui Li, et al., “PU-GAN: a Point Cloud Upsampling Adversarial Network”. ICCV2019 ! Shiyang Cheng, et al., “MeshGAN: Non-linear 3D Morphable Models of Faces”. arXiv: 1903.10384 ! Thomas Schlegl, et al., “Unsupervised Anomaly Detection with Generative Adversarial Networks to Guide Marker Discovery”. IPMI2017 ! Houssam Zenati, et al., “Efficient GAN-Based Anomaly Detection”. ICLRW2018 ! Dan Li, et al., “Anomaly Detection with Generative Adversarial Networks for Multivariate Time Series”. arXiv: 1809.04758 ! Pramuditha Perera, et al., “OCGAN: One-class Novelty Detection Using GANs with Constrained Latent Representations”. CVPR2019 ! Jesse Engel, et al., “GANSynth: Adversarial Neural Audio Synthesis”. ICLR2019 ! Chris Donahue, et al., “Adversarial Audio Synthesis”. ICLR2019
  54. 54. ! Andrés Marafioti, et al., “Adversarial Generation of Time-Frequency Features with application in audio synthesis”. ICML2019 ! Santiago Pascual, et al., “SEGAN: Speech Enhancement Generative Adversarial Network”. INTERSPEECH2017 ! Kou Tanaka, et al., “WaveCycleGAN: Synthetic-to-natural speech waveform conversion using cycle- consistent adversarial networks”. STL2018 ! Kou Tanaka, et al., “WaveCycleGAN2: Time-domain Neural Post-filter for Speech Waveform Generation”. arXiv: 1904.02892 ! Takuhiro Kaneko, et al., “CycleGAN-VC: Non-parallel Voice Conversion Using Cycle-Consistent Adversarial Networks”. EUSIPCO2018 ! Takuhiro Kaneko, et al., “CycleGAN-VC2: Improved CycleGAN-based Non-parallel Voice Conversion”. ICASSP2019 ! Hirokazu Kameoka, et al., “StarGAN-VC: Non-parallel many-to-many voice conversion with star generative adversarial networks”. arXiv: 1806.02169 ! Takuhiro Kaneko, et al., “StarGAN-VC2: Rethinking Conditional Methods for StarGAN-Based Voice Conversion”. INTERSPEECH2019 ! “AdaGAN: Adaptive GAN for Many-to-Many Non-Parallel Voice Conversion”. ICLR2020 under review ! Yuki Saito, et al., “Statistical Parametric Speech Synthesis Incorporating Generative Adversarial Networks”. IEEE/ACM Transactions on Audio, Speech, and Language Processing 2018 ! Mikołaj Bińkowski, et al., “High Fidelity Speech Synthesis with Adversarial Networks”. arXiv: 1909.11646 ! Ju-chieh Chou, et al., “One-shot Voice Conversion by Separating Speaker and Content Representations with Instance Normalization”. INTERSPEECH2019

×