Research map / Label-efficient and robust learning / Recognition and long-tail
Letter-Level Online Writer Identification
Zelin Chen, Hong-Xing Yu, Ancong Wu, Wei‐Shi Zheng
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 letters and styles adapter (LSA) (ablation) | full model · LERID (open-set) | rank-1 accuracy (%): 85.4 -> 93.8 | Improved |
| add letters and styles adapter (LSA) (ablation) | full model · Modified IAM (open-set) | rank-1 accuracy (%): 64.5 -> 76.2 | Improved |
| add letter-specific feature selection (ablation) | LSA · Modified IAM (open-set) | rank-1 accuracy (%): 75.2 -> 76.2 | Improved |
| use style-letter-specific adapters instead of a universal adapter (ablation) | LSA · Modified IAM (open-set) | rank-1 accuracy (%): 31.9 -> 76.2 | Improved |
| use style-letter-specific adapters instead of letter-specific only (ablation) | LSA · LERID (open-set) | rank-1 accuracy (%): 92.3 -> 93.8 | Improved |
| add hierarchical attention pooling (HAP) instead of mean pooling (ablation) | full model · Modified IAM (open-set) | rank-1 accuracy (%): 72.4 -> 76.2 | Improved |
| add hierarchical attention pooling (HAP) instead of max pooling (ablation) | full model · LERID (open-set) | rank-1 accuracy (%): 90.3 -> 93.8 | Improved |
| add temporal attention to style attention (ablation) | HAP · LERID (open-set) | rank-1 accuracy (%): 91.2 -> 92.7 | Improved |
| add letter attention (ablation) | HAP · LERID (open-set) | rank-1 accuracy (%): 92.7 -> 93.8 | Improved |
| place style hierarchy before temporal hierarchy (ablation) | HAP · LERID (open-set) | rank-1 accuracy (%): 93.6 -> 93.8 | Improved |
| add HAP without LSA (ablation) | model with max pooling · Modified IAM (open-set) | rank-1 accuracy (%): 68.5 -> 64.5 | Hurt |
| add LSA without HAP (ablation) | model with max pooling · Modified IAM (open-set) | rank-1 accuracy (%): 68.5 -> 67.8 | No clear effect |
Closest papers by meaning
- Few-Shot Detection of Machine-Generated Text using Style Representations (ICLR 2024 · similarity 0.83)
- MetaWriter: Personalized Handwritten Text Recognition Using Meta-Learned Prompt Tuning (CVPR 2025 · similarity 0.83)
- DeTeCtive: Detecting AI-generated Text via Multi-Level Contrastive Learning (NeurIPS 2024 · similarity 0.81)
- General Detection-based Text Line Recognition (NeurIPS 2024 · similarity 0.81)
- Towards Automatic Boundary Detection for Human-AI Collaborative Hybrid Essay in Education (AAAI 2024 · similarity 0.80)
- Human Texts Are Outliers: Detecting LLM-generated Texts via Out-of-distribution Detection (NeurIPS 2025 · similarity 0.80)
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.
