Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Mitigating Knowledge Discrepancies among Multiple Datasets for Task-agnostic Unified Face Alignment
Jiahao Xia, Min Gang Xu, Wenjian Huang, Jianguo Zhang, Haimin Zhang, Chunxia Xiao
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 |
|---|---|---|---|
| train on multiple datasets instead of a single dataset | TUFA (ViT-S/8) · WFLW (full set) | NME: 4.23 -> 3.93 | Improved |
| train on multiple datasets instead of a single dataset | TUFA (ViT-S/8) · 300W (full set) | NME: 3.23 -> 2.95 | Improved |
| train on multiple datasets instead of a single dataset | TUFA (ViT-S/8) · COFW | NME: 3.40 -> 3.07 | Improved |
| train on multiple datasets instead of a single dataset | TUFA (ViT-S/8) · COFW68 (cross-dataset) | NME: 4.11 -> 3.70 | Improved |
| train on multiple datasets instead of a single dataset | TUFA (ViT-S/8) · Masked 300W (full set) | NME: 7.67 -> 5.58 | Improved |
| train on multiple datasets instead of a single dataset | TUFA (ViT-S/8) · MERL-RAV (full set) | NME_box: 1.55 -> 1.54 | No clear effect |
| add COFW as a second training dataset to 300W (ablation) | TUFA (ViT-S/8) · 300W | NME: 3.23 -> 3.14 | Improved |
| add WFLW as a third training dataset (ablation) | TUFA (ViT-S/8) · 300W | NME: 3.14 -> 3.08 | Improved |
| add MERL-RAV as a fourth training dataset (ablation) | TUFA (ViT-S/8) · 300W | NME: 3.08 -> 2.95 | Improved |
| add semantic alignment embeddings to affine-aligned mean shapes (ablation) | TUFA (ViT-S/8) · 300W | NME: 3.42 -> 2.95 | Improved |
| add semantic alignment embeddings (ablation) | TUFA (ViT-S/8) · 300W-private | NME: 4.08 -> 3.40 | Improved |
| use unified plane with structure prompts instead of multi-task learning (ablation) | TUFA (ViT-S/8) · MERL-RAV | NME_box: 1.94 -> 1.54 | Improved |
Closest papers by meaning
- How Far Are We From Solving the 2D & 3D Face Alignment Problem? (And a Dataset of 230,000 3D Facial Landmarks) (ICCV 2017 · similarity 0.87)
- Faster Than Real-Time Facial Alignment: A 3D Spatial Transformer Network Approach in Unconstrained Poses (ICCV 2017 · similarity 0.87)
- DeCaFA: Deep Convolutional Cascade for Face Alignment in the Wild (ICCV 2019 · similarity 0.87)
- Semantic Alignment: Finding Semantically Consistent Ground-Truth for Facial Landmark Detection (CVPR 2019 · similarity 0.87)
- Face Alignment Across Large Poses: A 3D Solution (CVPR 2016 · similarity 0.87)
- Generalizable Face Landmarking Guided by Conditional Face Warping (CVPR 2024 · similarity 0.85)
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.
