Facial expression transfer using StyleGAN
Facial expression transfer is a hot topic in computer vision and computer graphics. Given a source image and a target expression, the goal is to transfer the target expression to the source image. StyleGAN, a notable improvement in the development of Generative Adversarial Networks (GANs), is a powe...
Saved in:
| 主要作者: | |
|---|---|
| 其他作者: | |
| 格式: | Final Year Project |
| 語言: | English |
| 出版: |
Nanyang Technological University
2023
|
| 主題: | |
| 在線閱讀: | https://hdl.handle.net/10356/167195 |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 機構: | Nanyang Technological University |
| 語言: | English |
| 總結: | Facial expression transfer is a hot topic in computer vision and computer graphics. Given a source image and a target expression, the goal is to transfer the target expression to the source image. StyleGAN, a notable improvement in the development of Generative Adversarial Networks (GANs), is a powerful tool to generate high-fidelity and high-resolution images of human faces. This project designs a pipeline to perform facial expression transfer with customized StyleGAN and Pixel2style2pixel (pSp) models. Specifically, pSp encoder is used to obtain the latent representation (latent code) of the source image, which is afterwards manipulated based on the target expression. StyleGAN generator is then used to generate the resulting image from the manipulated latent code. This project also explores the latent space of StyleGAN, analyses problems like model generalizability and expressiveness-editability trade-off. Finally, a pipeline is successfully built for both image and video-based facial expression transfer. Experiment results demonstrate the effectiveness of the proposed method. |
|---|