Research map / Label-efficient and robust learning / Recognition and long-tail
Recursive Spatial Transformer (ReST) for Alignment-Free Face Recognition
Wanglong Wu, Meina Kan, Xin Liu, Yi Yang, Shiguang Shan, Xilin Chen
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 one recursion (a single spatial transformer) (ablation) | AlexNet · LFW | mean accuracy (%): 97.37 -> 98.03 | Improved |
| increase recursions from 1 to 4 (ablation) | ReST on AlexNet · LFW | mean accuracy (%): 98.03 -> 98.38 | Improved |
| increase recursions from 4 to 5 (ablation) | ReST on AlexNet · LFW | mean accuracy (%): 98.38 -> 98.20 | Hurt |
| increase recursions from 0 to 4 (ablation) | ReST on AlexNet · LFW open-set | DIR (%) @ FAR=1%: 53.3 -> 65.4 | Improved |
| replace pre-alignment with ReST (alignment-free, 4 recursions) | AlexNet · LFW | mean accuracy (%): 98.06 -> 98.38 | Improved |
| add hierarchy with 3 regions (HiReST-3) to ReST (ablation) | ReST · LFW alignment-free | mean accuracy (%): 98.38 -> 98.62 | Improved |
| add 2-layer hierarchy with 9 regions (HiReST-9) to ReST (ablation) | ReST · LFW alignment-free | mean accuracy (%): 98.38 -> 98.90 | Improved |
| add 2-layer hierarchy (HiReST-9) to ReST (ablation) | ReST · LFW alignment-free | DIR (%) @ FAR=1%: 65.4 -> 77.2 | Improved |
| make HiReST-9 alignment-free with 4 recursions instead of pre-aligned | HiReST-9 · LFW open-set | DIR (%) @ FAR=1%: 63.2 -> 77.2 | Improved |
| make HiReST-9 alignment-free with 4 recursions instead of pre-aligned | HiReST-9 · LFW | mean accuracy (%): 98.70 -> 98.90 | Improved |
| remove pre-alignment from plain AlexNet (no ReST) (ablation) | AlexNet · LFW | mean accuracy (%): 98.06 -> 97.37 | Hurt |
Closest papers by meaning
- Recurrent Convolutional Shape Regression (TPAMI 2018 · similarity 0.89)
- A Recurrent Encoder-Decoder Network for Sequential Face Alignment (ECCV 2016 · similarity 0.88)
- Direct Shape Regression Networks for End-to-End Face Alignment (CVPR 2018 · similarity 0.88)
- Supervised Transformer Network for Efficient Face Detection (ECCV 2016 · similarity 0.88)
- Faster Than Real-Time Facial Alignment: A 3D Spatial Transformer Network Approach in Unconstrained Poses (ICCV 2017 · similarity 0.87)
- RED-Net: A Recurrent Encoder–Decoder Network for Video-Based Face Alignment (IJCV 2018 · 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.
