Research map / Architectures and efficiency / Spiking networks
Spike-HTR: Spiking Neural Transformer for Handwritten Text Recognition
Xiubo Liang, Jinxing Han, Yuke Li, Haoqi Zhu, Yu Zhao, Hongzhi Wang
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 |
|---|---|---|---|
| use InkCoder instead of repeat injection (ablation) | Spike-HTR · IAM val | validation CER (%): 4.3 -> 3.5 | Improved |
| use InkCoder instead of Poisson coding (ablation) | Spike-HTR · IAM val | validation CER (%): 4.7 -> 3.5 | Improved |
| use InkCoder instead of TTFS-style coding (ablation) | Spike-HTR · IAM val | validation CER (%): 4.1 -> 3.5 | Improved |
| fuse edge stream with intensity stream (ablation) | Spike-HTR · IAM val | validation CER (%): 3.6 -> 3.5 (intensity-only), 3.8 -> 3.5 (edge-only) | Improved |
| use gradual gate contraction over timesteps (ablation) | Spike-HTR · IAM val | validation CER (%): 3.7 -> 3.5 | Improved |
| use spiking neurons instead of GELU (ablation) | Spike-HTR · IAM val (T=1) | validation CER (%): 3.558 -> 3.489 | Improved |
| add InkCoder and membrane residual (ablation) | Spike-HTR · IAM val (T=1) | validation CER (%): 3.646 -> 3.489 | Improved |
| add keep-and-merge length reduction (ablation) | Spike-HTR · IAM val | validation CER (%): 3.52 -> 3.49 (r 1.00 -> 0.68) | No clear effect |
| keep-and-merge instead of merge-only (ablation) | Spike-HTR · IAM val | validation CER (%): 3.60 -> 3.49 | Improved |
| stop gradient through the preview (ablation) | Spike-HTR · IAM val | validation CER (%): 3.65 -> 3.49 | Improved |
| increase timesteps from T=2 to T=4 | Spike-HTR · IAM test | test CER (%): 5.4 -> 5.4 (IAM test CER) | No clear effect |
| increase timesteps | Spike-HTR · IAM, LAM, READ2016 | WER | Improved |
Closest papers by meaning
- MetaWriter: Personalized Handwritten Text Recognition Using Meta-Learned Prompt Tuning (CVPR 2025 · similarity 0.86)
- On the Generalization of Handwritten Text Recognition Models (CVPR 2025 · similarity 0.85)
- Spike2Former: Efficient Spiking Transformer for High-performance Image Segmentation (AAAI 2025 · similarity 0.85)
- Toward Relative Positional Encoding in Spiking Transformers (NeurIPS 2025 · similarity 0.84)
- Shrinking Your TimeStep: Towards Low-Latency Neuromorphic Object Recognition with Spiking Neural Networks (AAAI 2024 · similarity 0.83)
- Towards Efficient Spiking Transformer: a Token Sparsification Framework for Training and Inference Acceleration (ICML 2024 · similarity 0.83)
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.
