Research map / Image generation / Editing and style transfer
Semi-supervised Adversarial Learning to Generate Photorealistic Face Images of New Identities from 3D Morphable Model
Baris Gecer, Binod Bhattarai, Josef Kittler, Tae-Kyun Kim
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 preservation loss L_C (ablation) | Face generation framework (VGG 50% + GANFaces) · IJB-A verification (VGG 50% training data) | TAR @ FAR=0.01: 0.50532 -> 0.53507 | Improved |
| add identity preservation loss L_C (ablation) | Face generation framework (VGG 50% + GANFaces) · IJB-A verification (VGG 50% training data) | TAR @ FAR=0.001: 0.17636 -> 0.18768 | Improved |
| add adversarial pair matching L_DP (ablation) | Face generation framework (VGG 50% + GANFaces) · IJB-A verification (VGG 50% training data) | TAR @ FAR=0.01: 0.48034 -> 0.53507 | Improved |
| add adversarial pair matching L_DP (ablation) | Face generation framework (VGG 50% + GANFaces) · IJB-A verification (VGG 50% training data) | TAR @ FAR=0.001: 0.15300 -> 0.18768 | Improved |
| add cycle consistency loss L_cyc (ablation) | Face generation framework (VGG 50% + GANFaces) · IJB-A verification (VGG 50% training data) | TAR @ FAR=0.01: 0.49701 -> 0.53507 | Improved |
| add cycle consistency loss L_cyc (ablation) | Face generation framework (VGG 50% + GANFaces) · IJB-A verification (VGG 50% training data) | TAR @ FAR=0.001: 0.18341 -> 0.18768 | No clear effect |
| add GANFaces synthetic images to 50% of VGG Face training data | Face generation framework (VGG 50% + GANFaces) · IJB-A verification (VGG 50% training data) | TAR @ FAR=0.01: 0.49751 -> 0.53507 | Improved |
| add GANFaces synthetic images to 50% of VGG Face training data | Face generation framework (VGG 50% + GANFaces) · IJB-A verification (VGG 50% training data) | TAR @ FAR=0.001: 0.17580 -> 0.18768 | Improved |
| add GANFaces synthetic images to full VGG Face training data | VGG19 face recognition · LFW verification | accuracy (%): 94.8 -> 94.9 | Improved |
| add GANFaces synthetic images to full VGG Face training data | VGG19 face recognition · LFW verification | 100% - EER: 94.6 -> 95.1 | Improved |
Closest papers by meaning
- Towards Open-Set Identity Preserving Face Synthesis (CVPR 2018 · similarity 0.90)
- A 3D GAN for Improved Large-Pose Facial Recognition (CVPR 2021 · similarity 0.89)
- A Decoupled 3D Facial Shape Model by Adversarial Training (ICCV 2019 · similarity 0.88)
- Dual-Agent GANs for Photorealistic and Identity Preserving Profile Face Synthesis (NeurIPS 2017 · similarity 0.88)
- Data Synthesis with Diverse Styles for Face Recognition via 3DMM-Guided Diffusion (CVPR 2025 · similarity 0.87)
- Non-Adaptive Adversarial Face Generation (NeurIPS 2025 · 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.
