Research map / Label-efficient and robust learning / Recognition and long-tail
Multi-Directional Multi-Level Dual-Cross Patterns for Robust Face Recognition
Changxing Ding, Jonghyun Choi, Dacheng Tao, Larry Steven Davis
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 |
|---|---|---|---|
| combine the two dual-cross encoders (DCP-1 + DCP-2) into DCP (ablation) | DCP-1 / DCP-2 · FERET mean rank-1 | identification rate (%): 91.26 -> 92.80 | Improved |
| add multi-directional FDG gradient filtering to DCP (MD-DCPs) (ablation) | DCP · FERET hardest 146 probe images | identification rate (%): 17.12 -> 33.56 | Improved |
| add multi-directional FDG gradient filtering to DCP (MD-DCPs) (ablation) | DCP · CAS-PEAL-R1 PL | identification rate (%): 50.25 -> 65.23 | Improved |
| add multi-directional FDG gradient filtering to DCP (MD-DCPs) (ablation) | DCP · FRGC 2.0 Experiment 4 | mean verification rate (%): 22.62 -> 30.74 | Improved |
| use dual-cross grouping versus other 35 grouping modes (ablation) | DCP encoding · FERET gallery images | joint Shannon entropy | Improved |
| use WPCA instead of PLDA classifier with MDML-DCPs (ablation) | MDML-DCPs · CAS-PEAL-R1 PL | rank-1 identification rate (%): 63.26 -> 82.92 | Improved |
| use linear SVM score fusion instead of score averaging (ablation) | MDML-DCPs + PLDA · FRGC 2.0 Experiment 4 ROC III | verification rate (%): 92.89 -> 93.39 | Improved |
| use linear SVM score fusion instead of score averaging (ablation) | MDML-DCPs + PLDA · LFW View 2 unrestricted | accuracy (%): 94.57 -> 95.13 | Improved |
| fuse JB and PLDA classifiers (18 scores) with linear SVM (ablation) | MDML-DCPs · LFW View 2 unrestricted | accuracy (%): 95.13 -> 95.58 | Improved |
| use JB classifier instead of PLDA (ablation) | MDML-DCPs + linear SVM · LFW View 2 unrestricted | accuracy (%): 95.13 -> 95.40 | Improved |
| add flip trick to DCP | DCP · LFW View 2 restricted | accuracy (%): 75.00 -> 76.37 | Improved |
| use histogram intersection instead of chi-squared metric (ablation) | DCP · FERET mean rank-1 | identification rate (%): 92.80 -> 92.66 | No clear effect |
Closest papers by meaning
- DVG-Face: Dual Variational Generation for Heterogeneous Face Recognition (TPAMI 2021 · similarity 0.83)
- Disentangled Variational Representation for Heterogeneous Face Recognition (AAAI 2019 · similarity 0.83)
- DCFace: Synthetic Face Generation With Dual Condition Diffusion Model (CVPR 2023 · similarity 0.83)
- Towards Lightweight Pixel-Wise Hallucination for Heterogeneous Face Recognition (TPAMI 2022 · similarity 0.83)
- Coupled Deep Learning for Heterogeneous Face Recognition (AAAI 2018 · similarity 0.83)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.81)
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.
