Research map / Label-efficient and robust learning / Recognition and long-tail
Faceptor: A Generalist Model for Face Perception
Lixiong Qin, Mei Wang, Xuannan Liu, Yuhang Zhang, Wei Deng, Xiaoshuai Song, Weiran Xu, Weihong Deng
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 |
|---|---|---|---|
| replace per-category output heads (Naive Faceptor) with shared task-specific queries (Faceptor) | Faceptor-Base · parameters | parameters (M): 178.9 -> 103.2 | Improved |
| use task-specific queries instead of standardized heads | Faceptor-Base · RAF-DB | accuracy: 87.58 -> 90.38 | Improved |
| use task-specific queries instead of standardized heads | Faceptor-Base · Face recognition mean | 1:1 verification accuracy mean: 95.57 -> 95.10 | Hurt |
| add Layer-Attention trained with two-stage process (ablation) | Faceptor-Base · RAF-DB | accuracy: 89.80 -> 90.38 | Improved |
| add Layer-Attention trained with two-stage process (ablation) | Faceptor-Base · MORPH II | MAE: 1.882 -> 1.869 | Improved |
| add Layer-Attention directly without two-stage training (ablation) | Faceptor-Base · MORPH II | MAE: 1.882 -> 1.970 | Hurt |
| add Layer-Attention to binary attribute classification (ablation) | Faceptor-Base · CelebA | mAcc: 91.40 -> 91.43 | No clear effect |
| add auxiliary tasks (landmark, parsing, recognition) to age estimation (ablation) | Faceptor · MORPH II | MAE: 1.869 -> 1.787 | Improved |
| add auxiliary tasks to expression recognition (ablation) | Faceptor · RAF-DB | accuracy: 90.38 -> 91.92 | Improved |
| joint training on 13 datasets from FaRL initialization versus specialized ViT trained from scratch | Faceptor-Full · Face verification mean | mean accuracy: 96.46 -> 95.28 | Hurt |
| tune only task-specific queries (prompt tuning) vs all parameters on novel semantics (ablation) | Faceptor-Base · LFW-73 transfer | mAcc: 85.56 vs 87.95 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ADNet — ADNet: Leveraging Error-Bias Towards Normal Direction in Face Alignment (ICCV 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FaceXFormer: A Unified Transformer for Facial Analysis (ICCV 2025 · similarity 0.91)
- Pre-training Strategies and Datasets for Facial Representation Learning (ECCV 2022 · similarity 0.86)
- Visual Psychophysics for Making Face Recognition Algorithms More Explainable (ECCV 2018 · similarity 0.85)
- FaceBench: A Multi-View Multi-Level Facial Attribute VQA Dataset for Benchmarking Face Perception MLLMs (CVPR 2025 · similarity 0.85)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.84)
- FSRT: Facial Scene Representation Transformer for Face Reenactment from Factorized Appearance Head-pose and Facial Expression Features (CVPR 2024 · similarity 0.84)
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.
