Research map / Adversarial robustness and security / Certified and empirical robustness
Improving Robustness of Facial Landmark Detection by Defending Against Adversarial Attacks
Congcong Zhu, Xiaoqiang Li, Jide Li, Songmin Dai
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 sample-adaptive adversarial training to HG*1 | HG*1 · 300W Full | NME: 4.21 -> 3.38 | Improved |
| add SAAT to HG*1 | HG*1 · 300W Challenging | NME: 7.13 -> 5.10 | Improved |
| add SAAT to MDM | MDM · 300W Full | NME: 4.78 -> 3.98 | Improved |
| add SAAT to HG*2 | HG*2 · 300W Full | NME: 3.76 -> 3.29 | Improved |
| add SAAT to HG*1 on masked faces | HG*1 · Masked 300W Full | NME: 10.56 -> 7.58 | Improved |
| add SAAT to MDM on 300VW trained on 300W only | MDM · 300VW Cate-1 | NME: 8.74 -> 8.74 | Hurt |
| add adaptive beta instead of beta = 0 (ablation) | HG*1+SAAT · 300W Full | NME: 4.45 -> 3.38 | Improved |
| add adaptive beta instead of fixed beta = 1 (ablation) | HG*1+SAAT · 300W Full | NME: 4.04 -> 3.38 | Improved |
| use lambda = 0.1 instead of lambda = 0 (ablation) | HG*1+SAAT · 300W Full | NME: 4.13 -> 3.38 | Improved |
| use lambda = 0.1 instead of lambda = 0.5 (ablation) | HG*1+SAAT · 300W Full | NME: 3.43 -> 3.38 | Improved |
| add structure map S (ablation) | SAAT · 300W Full | NME: 3.78 -> 3.38 | Improved |
| add semantic reconstruction loss L_R (ablation) | SAAT · 300W Full | NME: 3.99 -> 3.38 | Improved |
Closest papers by meaning
- Facial Landmark Detection with Tweaked Convolutional Neural Networks (TPAMI 2017 · similarity 0.89)
- Style Aggregated Network for Facial Landmark Detection (CVPR 2018 · similarity 0.88)
- Semantic Alignment: Finding Semantically Consistent Ground-Truth for Facial Landmark Detection (CVPR 2019 · similarity 0.87)
- PriSM: Parsing and Style-Mixed Consistency for Unsupervised Domain Adaptation in Facial Landmark Detection (ECCV 2026 · similarity 0.87)
- Facial Landmark Detection: A Literature Survey (IJCV 2018 · similarity 0.87)
- Brute-Force Facial Landmark Analysis With a 140,000-Way Classifier (AAAI 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.
