Research map / Label-efficient and robust learning / Recognition and long-tail
Orthogonal Deep Features Decomposition for Age-Invariant Face Recognition
Yitong Wang, Dihong Gong, Zheng Zhou, Xing Ji, Hao Wang, Zhifeng Li, Wei Liu, Tong Zhang
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 |
|---|---|---|---|
| decompose features into radial age-related and angular identity-related components (OE-CNNs vs A-Softmax) (ablation) | A-Softmax CNN · MORPH Album 2 (public datasets) | rank-1 identification rate: 96.27% -> 97.46% | Improved |
| add feature decomposition (OE-CNNs vs A-Softmax) with CAF in training (ablation) | A-Softmax CNN · MORPH Album 2 (public datasets + CAF) | rank-1 identification rate: 96.59% -> 98.57% | Improved |
| add the CAF cross-age dataset to training (ablation) | OE-CNNs · MORPH Album 2 | rank-1 identification rate: 97.46% -> 98.57% | Improved |
| add the CAF cross-age dataset to training (ablation) | Softmax · MORPH Album 2 | rank-1 identification rate: 94.84% -> 95.49% | Improved |
| add the CAF cross-age dataset to training (ablation) | A-Softmax · MORPH Album 2 | rank-1 identification rate: 96.27% -> 96.59% | Improved |
| use A-Softmax instead of Softmax for identity supervision (ablation) | CNN baseline · MORPH Album 2 (public datasets) | rank-1 identification rate: 94.84% -> 96.27% | Improved |
| replace LF-CNNs with OE-CNNs | MORPH Album 2 protocol · MORPH Album 2 (10,000 subjects) | rank-1 identification rate: 97.51% -> 98.55% | Improved |
| replace AE-CNNs with OE-CNNs | MORPH Album 2 protocol · MORPH Album 2 (3,000 subjects) | rank-1 identification rate: 98.13% -> 98.67% | Improved |
| add feature decomposition (OE-CNNs vs A-Softmax) (ablation) | A-Softmax CNN · CACD-VS | accuracy: 98.7% -> 99.2% | Improved |
| add feature decomposition (OE-CNNs vs A-Softmax) (ablation) | A-Softmax CNN · CACD-VS | AUC: 99.5% -> 99.5% | No clear effect |
| add feature decomposition (OE-CNNs vs A-Softmax) (ablation) | A-Softmax CNN · FG-NET (MF1 small) | rank-1 identification rate: 46.77% -> 52.67% | Improved |
| add 3-patch ensemble to OE-CNNs (ablation) | OE-CNNs · FG-NET (MF1 small) | rank-1 identification rate: 52.67% -> 58.21% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SphereFace — SphereFace: Deep Hypersphere Embedding for Face Recognition (CVPR 2017) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Latent Factor Guided Convolutional Neural Networks for Age-Invariant Face Recognition (CVPR 2016 · similarity 0.94)
- Towards Age-Invariant Face Recognition (TPAMI 2020 · similarity 0.92)
- Look across Elapse: Disentangled Representation Learning and Photorealistic Cross-Age Face Synthesis for Age-Invariant Face Recognition (AAAI 2019 · similarity 0.91)
- Decorrelated Adversarial Learning for Age-Invariant Face Recognition (CVPR 2019 · similarity 0.91)
- When Age-Invariant Face Recognition Meets Face Age Synthesis: A Multi-Task Learning Framework and a New Benchmark (TPAMI 2022 · similarity 0.91)
- When Age-Invariant Face Recognition Meets Face Age Synthesis: A Multi-Task Learning Framework (CVPR 2021 · similarity 0.90)
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.
