Research map / Image generation / GANs
Synthesis of High-Quality Visible Faces from Polarimetric Thermal Faces using Generative Adversarial Networks
He Zhang, Benjamin S. Riggan, Shuowen Hu, Nathaniel J. Short, Vishal M. Patel
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 multi-level pooling to the single-stream dense-residual encoder-decoder (ablation) | DR-ED · Polarimetric thermal-visible (Protocol 1) | EER: 16.51% -> 15.67% | Improved |
| use dense-residual encoder-decoder instead of GAN-VFS network (ablation) | GAN-VFS · Polarimetric thermal-visible (Protocol 1) | EER: 22.45% -> 16.51% | Improved |
| add multi-level pooling (ablation) | DR-ED · Polarimetric thermal-visible (Protocol 1) | PSNR: 18.28 -> 18.80 | Improved |
| fuse Stokes images at input level instead of using S0 alone (ablation) | single-stream network · Polarimetric thermal-visible (Protocol 1) | AUC: 85.74% -> 92.55% | Improved |
| use multi-stream feature-level fusion instead of input-level fusion (ablation) | S-Polar-IF · Polarimetric thermal-visible (Protocol 1) | EER: 15.67% -> 11.78% | Improved |
| use feature-level fusion instead of output-level fusion (ablation) | M-Polar-OF · Polarimetric thermal-visible (Protocol 1) | AUC: 92.69% -> 96.03% | Improved |
| use output-level fusion instead of input-level fusion (ablation) | S-Polar-IF · Polarimetric thermal-visible (Protocol 1) | EER: 15.67% -> 15.90% | No clear effect |
| add adversarial loss to L2 loss (ablation) | L2 · Polarimetric thermal-visible (Protocol 1) | EER: 18.07% -> 13.23% | Improved |
| add adversarial loss to L2 loss (ablation) | L2 · Polarimetric thermal-visible (Protocol 1) | PSNR: 17.57 -> 17.33 | Hurt |
| add perceptual loss to L2-GAN (ablation) | L2-GAN · Polarimetric thermal-visible (Protocol 1) | PSNR: 17.33 -> 18.99 | Improved |
| add identity-preserving loss (ablation) | L2-GAN-P · Polarimetric thermal-visible (Protocol 1) | EER: 11.79% -> 11.78% | No clear effect |
| add identity-preserving loss (ablation) | L2-GAN-P · Polarimetric thermal-visible (Protocol 1) | PSNR: 18.99 -> 19.55 | Improved |
Closest papers by meaning
- Synthesizing Coupled 3D Face Modalities by Trunk-Branch Generative Adversarial Networks (ECCV 2020 · similarity 0.85)
- A 3D GAN for Improved Large-Pose Facial Recognition (CVPR 2021 · similarity 0.85)
- GANFIT: Generative Adversarial Network Fitting for High Fidelity 3D Face Reconstruction (CVPR 2019 · similarity 0.85)
- Dual-Agent GANs for Photorealistic and Identity Preserving Profile Face Synthesis (NeurIPS 2017 · similarity 0.85)
- POLAR: A Portrait OLAT Dataset and Generative Framework for Illumination-Aware Face Modeling (CVPR 2026 · similarity 0.85)
- Diffusion-driven GAN Inversion for Multi-Modal Face Image Generation (CVPR 2024 · similarity 0.85)
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.
