Research map / Image generation / GANs
Disentangled Representation Learning GAN for Pose-Invariant Face Recognition
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Train multi-image DR-GAN with more images per subject rather than n=1 (ablation) | DR-GAN · CFP Frontal-Profile | verification accuracy: 90.82 ± 0.28 -> 93.41 ± 1.17 | Helped |
| Train multi-image DR-GAN with more images per subject rather than n=1 (ablation) | DR-GAN · CFP Frontal-Frontal | verification accuracy: 97.13 ± 0.68 -> 97.84 ± 0.79 | Helped |
| Use learned coefficient fusion instead of averaging image scores (ablation) | Multi-image DR-GAN · Multi-PIE large-pose probes, nt=5 | identification rate: 95.8 -> 96.5 | Helped |
| Use learned coefficient fusion instead of averaged cosine distances (ablation) | DR-GAN · IJB-A identification @Rank-5 | identification rate: 93.2 ± 0.8 -> 94.7 ± 1.1 | Helped |
| Replace the adversarial loss with L2 loss (ablation) | DR-GAN · Multi-PIE Average | identification rate: 89.2 -> 78.3 | Hurt |
| Use the learned representation rather than synthetic images for recognition (ablation) | DR-GAN · CFP Frontal-Profile | verification accuracy: 91.02 ± 1.59 -> 93.41 ± 1.17 | Helped |
| Use DR-GAN rather than the strongest listed baseline | DR-GAN · Multi-PIE Average | identification rate: 83.3 -> 89.2 | Helped |
| Use DR-GAN rather than the strongest listed baseline | DR-GAN · CFP Frontal-Profile | verification accuracy: 91.97 ± 1.70 -> 93.41 ± 1.17 | Helped |
| Use DR-GAN fusion rather than the strongest listed baseline | DR-GAN · IJB-A identification @Rank-1 | identification rate: 85.2 ± 1.8 -> 85.5 ± 1.5 | No clear effect |
| Use DR-GAN fusion rather than the strongest listed baseline | DR-GAN · IJB-A verification @FAR=.001 | verification rate: 55.2 ± 3.2 -> 53.9 ± 4.3 | Hurt |
Closest papers by meaning
- A 3D GAN for Improved Large-Pose Facial Recognition (CVPR 2021 · similarity 0.90)
- Dual-Agent GANs for Photorealistic and Identity Preserving Profile Face Synthesis (NeurIPS 2017 · similarity 0.89)
- Towards Pose Invariant Face Recognition in the Wild (CVPR 2018 · similarity 0.89)
- Adversarial Pose Regression Network for Pose-Invariant Face Recognitions (AAAI 2021 · similarity 0.88)
- Multi-Task Adversarial Network for Disentangled Feature Learning (CVPR 2018 · similarity 0.87)
- IB-GAN: Disentangled Representation Learning with Information Bottleneck Generative Adversarial Networks (AAAI 2021 · 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.