An Enhanced GAN for Image Generation
Source abstract
Abstract
Generative adversarial networks (GANs) with gaming abilities have been widely applied in image generation. However, gamistic generators and discriminators may reduce the robustness of the obtained GANs in image generation under varying scenes. Enhancing the relation of hierarchical information in a generation network and enlarging differences of different network architectures can facilitate more structural information to improve the generation effect for image generation. In this paper, we propose an enhanced GAN via improving a generator for image generation (EIGGAN). EIGGAN applies a spatial attention to a generator to extract salient information to enhance the truthfulness of the generated images. Taking into relation the context account, parallel residual operations are fused into a generation network to extract more structural information from the different layers. Finally, a mixed loss function in a GAN is exploited to make a tradeoff between speed and accuracy to generate more realistic images. Experimental results show that the proposed method is superior to popular methods, i.e., Wasserstein GAN with gradient penalty (WGAN-GP) in terms of many indexes, i.e., Frechet Inception Distance, Learned Perceptual Image Patch Similarity, Multi-Scale Structural Similarity Index Measure, Kernel Inception Distance, Number of Statistically-Different Bins, Inception Score and some visual images for image generation.
Source: openalex_inverted_index · Quality: baseline
Research context
OpenAlex topics: AI in cancer detection · Image Processing and 3D Reconstruction · Medical Image Segmentation Techniques
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
References (42)
Show source-provided references
1. Human action image generation with differential privacy
2. Virtual face image generation for illumination and pose insensitive face recognition
3. Face recognition based on fitting a 3D morphable model10.1109/TPAMI.2003.1227983 ↗
4. Generative adversarial networks10.1145/3422622 ↗
5. Adversarial latent autoencoders
6. Designing an encoder for stylegan image manipulation
7. ReStyle: A residual-based stylegan encoder via iterative refinement
8. C. Yu and W. Wang, “Diverse similarity encoder for deep GAN inversion,” arXiv preprint arXiv: 2108.10201, 2021.
9. T. Wei et al., “E2Style: Improve the efficiency and effectiveness of StyleGAN inversion,” arXiv preprint arXiv:2104.07661, 2022.10.1109/TIP.2022.3167305 ↗
10. Y. Xu, Y. Shen, J. Zhu, C. Yang, and B. Zhou, “Generative hierarchical features from synthesizing images,” arXiv preprint arXiv:2007.10379, 2021.10.1109/CVPR46437.2021.00441 ↗
11. Gans trained by a two time-scale update rule converge to a local nash equilibrium
12. The unreasonable effectiveness of deep features as a perceptual metric
13. Multiscale structural similarity for image quality assessment
14. M. Bińkowski, D. J. Sutherland, M. Arbel, and A. Gretton, “Demystifying MMD GANs,” arXiv preprint arXiv, abs/1801.01401, 2018.
15. E. Richardson and Y. Weiss, “On GANs and GMMs,” arXiv preprint arXiv:1805.12462, 2018.
16. Improved techniques for training GANs
17. H. Tang, D. Xu, H. Liu, and N. Sebe, “Asymmetric generative adversarial networks for image-to-image translation,” arXiv preprint arXiv:1912.06931, 2019.10.1109/IJCNN.2019.8851881 ↗
18. High-resolution face swapping via latent semantics disentanglement
19. Improving GAN training via feature space shrinkage
20. Gan-supervised dense visual alignment
21. A style-based generator architecture for generative adversarial networks
22. Hyperinverter: Improving stylegan inversion via hypernetwork
23. Hyperstyle: Stylegan inversion with hypernetworks for real image editing
24. D. Lee, J. Y. Lee, D. Kim, J. Choi, and J. Kim, “Fix the noise: Disentangling source feature for transfer learning of StyleGAN,” arXiv preprint arXiv:2204.14079, 2022.10.1109/CVPR52729.2023.01367 ↗
25. Noisytwins: Class-consistent and diverse image generation through stylegans
26. Progressive growing of gans for improved quality, stability, and variation
27. Generative adversarial nets
28. I. Goodfellow, “NIPS 2016 tutorial: Generative adversarial networks,” arXiv preprint arXiv:1406.2661, 2016.
29. Which training methods for gans do actually converge?
30. CBAM: Convolutional block attention module
31. Deep learning face attributes in the wild
32. A. Krizhevsk and G. Hinton, “Learning multiple layers of features from tiny images,” Handbook Syst.Autoimmu. Dis., vol. 1, no. 4, 2009.
33. D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014.
34. Very deep convolutional networks for large-scale image recognition
35. Unsupervised representation learning with deep convolutional generative adversarial networks
36. Improved training of wasserstein gans
37. Generative modeling by estimating gradients of the data distribution
38. Analyzing and improving the image quality of stylegan
39. TransGAN: Two pure transformers can make one strong gan, and that can scale up
40. ViTGAN: Training gans with vision transformers
41. Dual discriminator weighted mixture generative adversarial network for image generation10.1007/s12652-021-03667-y ↗
42. PFGAN: Fast transformers for image synthesis10.1016/j.patrec.2023.04.013 ↗