Research map / Image generation / Editing and style transfer
Adversarial Cross-Spectral Face Completion for NIR-VIS Face Recognition
Ran He, Jie Cao, Lingxiao Song, Zhenan Sun, Tieniu Tan
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 |
|---|---|---|---|
| synthesize VIS faces with CFC before matching with Light CNN | Light CNN · CASIA NIR-VIS 2.0 (10-fold) | Rank-1 accuracy: 96.7% -> 98.6% | Improved |
| fuse features of original NIR image and synthesized VIS image (CFC-Fuse) | CFC · CASIA NIR-VIS 2.0 (10-fold) | Rank-1 accuracy: 98.6 -> 99.5 | Improved |
| add CFC synthesis | Light CNN · CASIA NIR-VIS 2.0 (first fold) | Rank-1 accuracy: 96.84 -> 99.21 | Improved |
| add CFC synthesis | Light CNN · BUAA-VisNir | Rank-1 accuracy: 96.5% to 99.7% | Improved |
| add CFC synthesis | Light CNN · BUAA-VisNir | VR@FAR=0.1%: 86.7% to 97.8% | Improved |
| add CFC synthesis | Light CNN · Oulu-CASIA NIR-VIS | Rank-1 accuracy: 96.7 -> 99.9 | Improved |
| add CFC synthesis | Light CNN · Oulu-CASIA NIR-VIS | VR@FAR=0.1%: 65.1 -> 90.7 | Improved |
| use CycleGAN synthesis instead of direct matching | Light CNN · CASIA NIR-VIS 2.0 (first fold) | Rank-1 accuracy: 96.84 -> 87.23 | Hurt |
| use paired pose-corrected one-to-one synthesis (CFC) instead of CycleGAN | synthesis · CASIA NIR-VIS 2.0 (first fold) | Rank-1 accuracy: 87.23 -> 99.21 | Improved |
| add discriminator on final RGB face (L_GF, L_DF) (ablation) | CFC · CASIA NIR-VIS 2.0 (256x256 qualitative) | visual quality | Improved |
| add wavelet-based multi-scale discriminator instead of single-scale (ablation) | CFC · CASIA NIR-VIS 2.0 (256x256 qualitative) | visual quality | Improved |
| add identity preserving perceptual loss L_p (ablation) | CFC · CASIA NIR-VIS 2.0 (256x256 qualitative) | visual quality | Improved |
Closest papers by meaning
- Physically-Based Face Rendering for NIR-VIS Face Recognition (NeurIPS 2022 · similarity 0.92)
- Learning Invariant Deep Representation for NIR-VIS Face Recognition (AAAI 2017 · similarity 0.91)
- Not Afraid of the Dark: NIR-VIS Face Recognition via Cross-Spectral Hallucination and Low-Rank Embedding (CVPR 2017 · similarity 0.91)
- LAMP-HQ: A Large-Scale Multi-pose High-Quality Database and Benchmark for NIR-VIS Face Recognition (IJCV 2021 · similarity 0.91)
- Multi-Energy Guided Image Translation with Stochastic Differential Equations for Near-Infrared Facial Expression Recognition (AAAI 2024 · similarity 0.87)
- Hypergraph-Guided Disentangled Spectrum Transformer Networks for Near-Infrared Facial Expression Recognition (AAAI 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.
