Research map / 3D vision / Face and shape
Fast-GANFIT: Generative Adversarial Network for High Fidelity 3D Face Reconstruction
Barış Geçer, Stylianos Ploumpis, Irene Kotsia, 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 |
|---|---|---|---|
| add identity losses L_id, L_id-hat and L_con together (ablation) | GANFit · MICC cooperative | shape error (mm): 1.20 -> 0.95 | Improved |
| add L_id-hat (ablation) | GANFit · MICC cooperative | shape error (mm): 1.07 -> 0.95 | Improved |
| add L_id alone (ablation) | GANFit · MICC cooperative | shape error (mm): 0.96 -> 0.95 | No clear effect |
| add L_con (ablation) | GANFit · MICC cooperative | shape error (mm): 0.94 -> 0.95 | No clear effect |
| add pixel loss L_pix (ablation) | GANFit · MICC cooperative | shape error (mm): 0.94 -> 0.95 | No clear effect |
| add landmark loss L_lan (ablation) | GANFit · MICC cooperative | shape error (mm): 0.91 -> 0.95 | Hurt |
| add shape regularizer Reg(p_s) (ablation) | GANFit · MICC cooperative | shape error (mm): 1.03 -> 0.95 | Improved |
| add expression regularizer Reg(p_e) (ablation) | GANFit · MICC cooperative | shape error (mm): 0.95 -> 0.95 | No clear effect |
| add Reg(p_l) lighting regularizer (ablation) | GANFit · MICC cooperative | shape error (mm): 0.96 -> 0.95 | No clear effect |
| initialise GANFit with FastGANFit regression (GANFit++) | GANFit · MICC outdoor | shape error (mm): 0.95 -> 0.94 | No clear effect |
| replace VAE texture model with GAN | texture model · UV texture, Ours (9358) | FID: 120.44 -> 2.95 | Improved |
| replace PCA texture model with GAN | texture model · UV texture, Ours (9358) | FID: 65.80 -> 2.95 | Improved |
Later papers that compared with this method
1 later accepted paper reports Fast-GANFit as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- FitMe: Deep Photorealistic 3D Morphable Model Avatars (CVPR 2023 · better on 3 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- GANFIT: Generative Adversarial Network Fitting for High Fidelity 3D Face Reconstruction (CVPR 2019 · similarity 0.98)
- Unsupervised High-Fidelity Facial Texture Generation and Reconstruction (ECCV 2022 · similarity 0.92)
- Towards High-Fidelity Face Self-Occlusion Recovery via Multi-View Residual-Based GAN Inversion (AAAI 2022 · similarity 0.92)
- Towards High-Fidelity 3D Face Reconstruction From In-the-Wild Images Using Graph Convolutional Networks (CVPR 2020 · similarity 0.91)
- 3DFaceGAN: Adversarial Nets for 3D Face Representation, Generation, and Translation (IJCV 2020 · similarity 0.91)
- Diffusion-driven GAN Inversion for Multi-Modal Face Image Generation (CVPR 2024 · similarity 0.88)
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.
