Research map / 3D vision / Face and shape
FSGANv2: Improved Subject Agnostic Face Swapping and Reenactment
Yuval Nirkin, Yosi Keller, Tal Hassner
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 face landmarks and flipped source image to the inpainting network Gc (ablation) | Gc · FaceForensics++ face inpainting | FID: 0.74 -> 0.62 | Improved |
| add face landmarks and flipped source image to Gc (ablation) | Gc · FaceForensics++ face inpainting | L1: 0.31 -> 0.16 | Improved |
| add face landmarks and flipped source image to Gc (ablation) | Gc · FaceForensics++ face inpainting | landmarks error: 19.4 -> 17.9 | Improved |
| add face landmarks and flipped source image to Gc (ablation) | Gc · FaceForensics++ face inpainting | id: 0.64 ± 0.11 -> 0.64 ± 0.10 | No clear effect |
| replace previous FSGAN inpainting with new Gc with landmarks (ablation) | FSGAN inpainting · FaceForensics++ face inpainting | FID: 2.64 -> 0.62 | Improved |
| replace previous FSGAN reenactment generator with new landmark-based Gr (ablation) | reenactment generator · FaceForensics++ face swapping | L1: 0.26 -> 0.22 | Improved |
| add completion network Gc to Gr (ablation) | FSGANv2 pipeline · FaceForensics++ face swapping | landmarks error: 31.8 ± 5.66 -> 23.2 ± 4.95 | Improved |
| add blending network Gb to Gr (ablation) | FSGANv2 pipeline · FaceForensics++ face swapping | FID: 0.71 -> 0.54 | Improved |
| add blending network Gb to Gr (ablation) | FSGANv2 pipeline · FaceForensics++ face swapping | L1: 0.22 -> 0.20 | Improved |
| add completion network Gc to Gr+Gb (full pipeline) (ablation) | FSGANv2 pipeline · FaceForensics++ face swapping | landmarks error: 28.0 ± 5.17 -> 21.6 ± 5.25 | Improved |
| add completion network Gc to Gr+Gb (full pipeline) (ablation) | FSGANv2 pipeline · FaceForensics++ face swapping | id: 0.41 ± 0.06 -> 0.37 ± 0.06 | Hurt |
| fine-tune on FaceForensics++ source images | FSGANv2 · FaceForensics++ face swapping | id: 0.37 ± 0.05 -> 0.44 ± 0.04 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FSGAN — FSGAN: Subject Agnostic Face Swapping and Reenactment (ICCV 2019) · better on 2 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
- FSGAN: Subject Agnostic Face Swapping and Reenactment (ICCV 2019 · similarity 0.98)
- Fine-Grained Face Swapping via Regional GAN Inversion (CVPR 2023 · similarity 0.90)
- One-Shot Neural Face Reenactment via Finding Directions in GAN’s Latent Space (IJCV 2024 · similarity 0.90)
- High-Resolution Face Swapping via Latent Semantics Disentanglement (CVPR 2022 · similarity 0.89)
- Fast Face-Swap Using Convolutional Neural Networks (ICCV 2017 · similarity 0.88)
- In-N-Out: Faithful 3D GAN Inversion with Volumetric Decomposition for Face Editing (CVPR 2024 · similarity 0.86)
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.
