Unsupervised Image Translation Using Multi-Scale Residual GAN
Image translation is a classic problem of image processing and computer vision for transforming an image from one domain to another by learning the mapping between an input image and an output image. A novel Multi-scale Residual Generative Adversarial Network (MRGAN) based on unsupervised learning i...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-11-01
|
| Series: | Mathematics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2227-7390/10/22/4347 |
| _version_ | 1797464602525564928 |
|---|---|
| author | Yifei Zhang Weipeng Li Daling Wang Shi Feng |
| author_facet | Yifei Zhang Weipeng Li Daling Wang Shi Feng |
| author_sort | Yifei Zhang |
| collection | DOAJ |
| description | Image translation is a classic problem of image processing and computer vision for transforming an image from one domain to another by learning the mapping between an input image and an output image. A novel Multi-scale Residual Generative Adversarial Network (MRGAN) based on unsupervised learning is proposed in this paper for transforming images between different domains using unpaired data. In the model, a dual generater architecture is used to eliminate the dependence on paired training samples and introduce a multi-scale layered residual network in generators for reducing semantic loss of images in the process of encoding. The Wasserstein GAN architecture with gradient penalty (WGAN-GP) is employed in the discriminator to optimize the training process and speed up the network convergence. Comparative experiments on several image translation tasks over style transfers and object migrations show that the proposed MRGAN outperforms strong baseline models by large margins. |
| first_indexed | 2024-03-09T18:10:31Z |
| format | Article |
| id | doaj.art-0f4d17e1320541dfa751451907b9d140 |
| institution | Directory Open Access Journal |
| issn | 2227-7390 |
| language | English |
| last_indexed | 2024-03-09T18:10:31Z |
| publishDate | 2022-11-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Mathematics |
| spelling | doaj.art-0f4d17e1320541dfa751451907b9d1402023-11-24T09:10:06ZengMDPI AGMathematics2227-73902022-11-011022434710.3390/math10224347Unsupervised Image Translation Using Multi-Scale Residual GANYifei Zhang0Weipeng Li1Daling Wang2Shi Feng3School of Computer Science and Engineering, Northeastern University, Shenyang 110169, ChinaSchool of Computer Science and Engineering, Northeastern University, Shenyang 110169, ChinaSchool of Computer Science and Engineering, Northeastern University, Shenyang 110169, ChinaSchool of Computer Science and Engineering, Northeastern University, Shenyang 110169, ChinaImage translation is a classic problem of image processing and computer vision for transforming an image from one domain to another by learning the mapping between an input image and an output image. A novel Multi-scale Residual Generative Adversarial Network (MRGAN) based on unsupervised learning is proposed in this paper for transforming images between different domains using unpaired data. In the model, a dual generater architecture is used to eliminate the dependence on paired training samples and introduce a multi-scale layered residual network in generators for reducing semantic loss of images in the process of encoding. The Wasserstein GAN architecture with gradient penalty (WGAN-GP) is employed in the discriminator to optimize the training process and speed up the network convergence. Comparative experiments on several image translation tasks over style transfers and object migrations show that the proposed MRGAN outperforms strong baseline models by large margins.https://www.mdpi.com/2227-7390/10/22/4347image translationgenerative adversarial networkunsupervised learningobject migrationmulti-scale residual network |
| spellingShingle | Yifei Zhang Weipeng Li Daling Wang Shi Feng Unsupervised Image Translation Using Multi-Scale Residual GAN Mathematics image translation generative adversarial network unsupervised learning object migration multi-scale residual network |
| title | Unsupervised Image Translation Using Multi-Scale Residual GAN |
| title_full | Unsupervised Image Translation Using Multi-Scale Residual GAN |
| title_fullStr | Unsupervised Image Translation Using Multi-Scale Residual GAN |
| title_full_unstemmed | Unsupervised Image Translation Using Multi-Scale Residual GAN |
| title_short | Unsupervised Image Translation Using Multi-Scale Residual GAN |
| title_sort | unsupervised image translation using multi scale residual gan |
| topic | image translation generative adversarial network unsupervised learning object migration multi-scale residual network |
| url | https://www.mdpi.com/2227-7390/10/22/4347 |
| work_keys_str_mv | AT yifeizhang unsupervisedimagetranslationusingmultiscaleresidualgan AT weipengli unsupervisedimagetranslationusingmultiscaleresidualgan AT dalingwang unsupervisedimagetranslationusingmultiscaleresidualgan AT shifeng unsupervisedimagetranslationusingmultiscaleresidualgan |