affritical Intelligent thuat toan Gan cho AI

abc

1. A Style-Based Generator Architecture for Generative Adversarial Networks Presenter: PHAM XUAN HIEN - 2021198001 Chungbuk National University , Korea

2. Content 1. Introduction 2. Gan Overview 3. Some Experiments 4. My Experiments

3. Introduction Image-to-Image Translation with Conditional Adversarial Nets Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, Alexei A. Efros

4. GANs Overview

5. Generator Overview Generator very much inspired by U-Net

6. Discriminator Overview Generator very much inspired by U-Net

7. Some Examples

8. My experiment Discriminator Loss

9. Dataset Piccaso Dataset on Kaggle w ebsite

10. Prepare the dataset

11. Training Batch size=5 Epoch=200

12. Results Models

13. Results Epoch 1 Input Image Real Image Synthesized Image

14. Results Epoch 50 Input Image Synthesized Image Real Image

15. Results Epoch 100 Input Image Synthesized Image Real Image

16. Results Epoch 150 Input Image Synthesized Image Real Image

17. Results Epoch 200 Input Image Synthesized Image Real Image

18. Thank you for your attention.