Research map / Image generation / Editing and style transfer
SliderGAN: Synthesizing Expressive Face Images by Sliding 3D Blendshape Parameters
Evangelos Ververas, Stefanos P. Zafeiriou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace WGP-GP adversarial loss with relativistic discriminator (RaD) (ablation) | SliderGAN · 4DFAB expression transfer | IED: 7.932-03 -> 6.84e-03 | Improved |
| replace WGP-GP adversarial loss with relativistic discriminator (RaD) (ablation) | SliderGAN · LRW expression and speech synthesis | IED: 1.14e-02 -> 9.35e-03 | Improved |
| replace WGP-GP adversarial loss with relativistic discriminator (RaD) (ablation) | SliderGAN · EmotionNet expression recognition | average accuracy: 0.614 -> 0.636 | Improved |
| use 3D blendshape parameters instead of action unit activations | SliderGAN vs GANimation · 4DFAB expression transfer | IED: 1.04e-02 -> 6.84e-03 | Improved |
| use blendshape parameters instead of action units | SliderGAN vs GANimation · LRW expression and speech synthesis | IED: 3.07e-02 -> 9.35e-03 | Improved |
| use blendshape parameters instead of action units | SliderGAN vs GANimation · EmotionNet expression recognition | average accuracy: 0.579 -> 0.636 | Improved |
| use blendshape parameters instead of action units (discriminator regression) | SliderGAN vs GANimation · EmotionNet parameter regression vs fitting | relative parameter error: 0.427 -> 0.131 | Improved |
| use blendshape parameters instead of action units (discriminator regression) | SliderGAN vs GANimation · EmotionNet parameter consistency after translation | relative parameter error: 0.513 -> 0.258 | Improved |
| add identity loss L_id to the generator loss (ablation) | SliderGAN · face generation (qualitative) | image realism and identity | Improved |
| add semi-supervised generation loss L_gen on paired data (ablation) | SliderGAN · face generation (qualitative) | image quality | Improved |
Closest papers by meaning
- StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation (CVPR 2018 · similarity 0.87)
- BlendGAN: Implicitly GAN Blending for Arbitrary Stylized Face Generation (NeurIPS 2021 · similarity 0.87)
- TransferI2I: Transfer Learning for Image-to-Image Translation From Small Datasets (ICCV 2021 · similarity 0.87)
- DeepI2I: Enabling Deep Hierarchical Image-to-Image Translation by Transferring from GANs (NeurIPS 2020 · similarity 0.87)
- Controllable Image-to-Video Translation: A Case Study on Facial Expression Generation (AAAI 2019 · similarity 0.87)
- 3DFaceGAN: Adversarial Nets for 3D Face Representation, Generation, and Translation (IJCV 2020 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
