Research map / Label-efficient and robust learning / Recognition and long-tail
When Counting Meets HMER: Counting-Aware Network for Handwritten Mathematical Expression Recognition
Bohan Li, Ye Yuan, Dingkang Liang, Xiao Liu, Zhilong Ji, Jinfeng Bai, Wenyu Liu, Xiang Bai
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 positional encoding (ablation) | DWAP · CROHME 2014 | ExpRate: 51.48 -> 51.88 | Improved |
| add joint optimization with symbol counting (ablation) | DWAP + positional encoding · CROHME 2014 | ExpRate: 51.88 -> 55.23 | Improved |
| add counting vector to the decoder (ablation) | DWAP + positional encoding + joint optimization · CROHME 2014 | ExpRate: 55.23 -> 57.00 | Improved |
| add counting vector to the decoder (ablation) | DWAP + positional encoding + joint optimization · CROHME 2019 | ExpRate: 53.13 -> 54.88 | Improved |
| use both 3x3 and 5x5 kernels instead of 3x3 in counting module (ablation) | CAN-DWAP · CROHME 2014 | ExpRate: 54.92 -> 57.00 | Improved |
| use both 3x3 and 5x5 kernels instead of 5x5 in counting module (ablation) | CAN-DWAP · CROHME 2014 | ExpRate: 55.53 -> 57.00 | Improved |
| feed ground-truth counting vector to the decoder (ablation) | CAN-DWAP · CROHME 2014 | ExpRate: 57.00 -> 62.44 | Improved |
| add random disturbance to counting vector (more inaccurate) (ablation) | CAN-DWAP with counting GT · CROHME 2014 | ExpRate: 62.44 -> 60.10 -> 58.28 | Hurt |
| train counting jointly with HMER instead of counting alone (ablation) | CAN counting module · CROHME 2014 | MAE_Ave: 0.048 -> 0.033 | Improved |
| add counting module (CAN) to DWAP | DWAP · CROHME 2014 (no augmentation) | ExpRate: 51.48 -> 57.00 | Improved |
| add counting module (CAN) to DWAP with data augmentation | DWAP · CROHME 2014 (with augmentation) | ExpRate: 57.97 -> 65.58 | Improved |
| add counting module (CAN) to ABM | ABM · HME100K | ExpRate: 65.93 -> 68.09 | Improved |
Closest papers by meaning
- SSAN: A Symbol Spatial-Aware Network for Handwritten Mathematical Expression Recognition (AAAI 2025 · similarity 0.91)
- Uni-MuMER: Unified Multi-Task Fine-Tuning of Vision-Language Model for Handwritten Mathematical Expression Recognition (NeurIPS 2025 · similarity 0.89)
- TAMER: Tree-Aware Transformer for Handwritten Mathematical Expression Recognition (AAAI 2025 · similarity 0.88)
- Generating Handwritten Mathematical Expressions From Symbol Graphs: An End-to-End Pipeline (CVPR 2024 · similarity 0.86)
- Counting in Small Transformers: The Delicate Interplay between Attention and Feed-Forward Layers (ICML 2025 · similarity 0.83)
- Vision Transformer Off-the-Shelf: A Surprising Baseline for Few-Shot Class-Agnostic Counting (AAAI 2024 · 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.
