Research map / 3D vision / Human pose estimation
Hierarchical Attention-Based Age Estimation and Bias Analysis
Shakediel Hiba, Yosi Keller
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 Transformer-Encoder augmentation aggregation (ablation) | hierarchical age estimator (VGG16) · CACD (SE protocol) | MAE: 5.80 -> 5.35 | Improved |
| use 10 augmentations aggregated by Transformer-Encoder instead of single augmentation (ablation) | age estimator · MORPH II (SE protocol) | MAE: 2.63 -> 2.53 | Improved |
| use Transformer-Encoder instead of average-pool for 10 augmentations (ablation) | age estimator · MORPH II (SE protocol) | MAE: 2.58 -> 2.53 | Improved |
| increase augmentations from 10 to 15 (ablation) | age estimator · MORPH II (SE protocol) | MAE: 2.53 -> 2.60 | Hurt |
| use 8 encoder layers and 8 heads instead of 4 and 4 (ablation) | age estimator · MORPH II (SE protocol) | MAE: 2.53 -> 2.57 | Hurt |
| use 2 encoder layers and 2 heads instead of 4 and 4 (ablation) | age estimator · MORPH II (SE protocol) | MAE: 2.53 -> 2.55 | Hurt |
| use age bin size 1 instead of 10 (ablation) | age estimator · MORPH II (SE protocol) | MAE: 2.56 -> 2.53 | Improved |
| use age bin size 1 instead of 5 (ablation) | age estimator · MORPH II (SE protocol) | MAE: 2.56 -> 2.53 | Improved |
| add RandAugment learned augmentations | age estimator · MORPH II | MAE | No clear effect |
| train on gender-balanced MORPH II subset | age estimator VGG16 · MORPH II (SE protocol, gender-balanced) | MAE: decreased by ~0.95 years for men, ~0.35 years for women | Hurt |
| add SAE (5 augmentations, Transformer-Encoder 1 layer 2 heads) to shallow network | Shallow CNN · CIFAR-10 | accuracy: 56% -> 70% | Improved |
| add SAE with 2 encoder layers to VGG16 | VGG16 · CIFAR-10 | accuracy: 87 -> 92 | Improved |
Closest papers by meaning
- Deep Regression Forests for Age Estimation (CVPR 2018 · similarity 0.90)
- Using Ranking-CNN for Age Estimation (CVPR 2017 · similarity 0.89)
- Hierarchical Face Aging through Disentangled Latent Characteristics (ECCV 2020 · similarity 0.88)
- Expression-Invariant Age Estimation Using Structured Learning (TPAMI 2017 · similarity 0.88)
- Efficient Group-n Encoding and Decoding for Facial Age Estimation (TPAMI 2017 · similarity 0.88)
- A Call to Reflect on Evaluation Practices for Age Estimation: Comparative Analysis of the State-of-the-Art and a Unified Benchmark (CVPR 2024 · similarity 0.88)
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.
