Research map / Label-efficient and robust learning / Recognition and long-tail
Feature-Level Frankenstein: Eliminating Variations for Discriminative 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 |
|---|---|---|---|
| Use FLF to learn a representation invariant to source-domain background | digit recognition model · DIGITS tested on MNIST | accuracy (y | d): 99.2% -> 99.5% | Helped |
| Use FLF to remove lighting-condition information from the recognition representation | face recognition model · Extended YaleB | accuracy (y | d): 86.6% -> 90.1% | Helped |
| Use FLF for makeup-invariant face recognition | face recognition model · PR2017 | rank-1 accuracy: 92.3% -> 94.6% | Helped |
| Use FLF for makeup-invariant face recognition | face recognition model · TCSVT2014 | rank-1 accuracy: 94.8% -> 96.2% | Helped |
| Use FLF for makeup-invariant face recognition | face recognition model · FAM | rank-1 accuracy: 82.6% -> 91.4% | Helped |
| Use FLF with 18 dispelled attributes | CelebA face recognition · CelebA | face recognition accuracy: 85.4% -> 92.7% | Helped |
| Use FLF to learn face-attribute representations | face attribute recognition model · CelebA | face attribute recognition accuracy: 91.81% -> 93.26% | Helped |
| Use FLF to learn face-attribute representations | face attribute recognition model · LFWA | face attribute recognition accuracy: 85.28% -> 87.82% | Helped |
| Use FLF as a representation-learning step | DFW face recognition model · DFW | Genuine Acceptance Rate @1%FAR: 33.76% -> 51.78% | Helped |
| Use FLF as a representation-learning step | DFW face recognition model · DFW | Genuine Acceptance Rate @0.1%FAR: 17.73% -> 38.64% | Helped |
| Add FLF pretraining to MIRA | DFW face recognition model · DFW | Genuine Acceptance Rate @1%FAR: 89.04% -> 91.30% | Helped |
| Add FLF pretraining to MIRA | DFW face recognition model · DFW | Genuine Acceptance Rate @0.1%FAR: 75.08% -> 78.55% | Helped |
Closest papers by meaning
- Multi-Task Adversarial Network for Disentangled Feature Learning (CVPR 2018 · similarity 0.86)
- Masked Face Recognition with Generative-to-Discriminative Representations (ICML 2024 · similarity 0.85)
- Controllable Invariance through Adversarial Feature Learning (NeurIPS 2017 · similarity 0.85)
- Disentangling factors of variation in deep representation using adversarial training (NeurIPS 2016 · similarity 0.84)
- Kernel Learning with Adversarial Features: Numerical Efficiency and Adaptive Regularization (NeurIPS 2025 · similarity 0.83)
- Rethinking Invariance Regularization in Adversarial Training to Improve Robustness-Accuracy Trade-off (ICLR 2025 · similarity 0.83)
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.