Research map / Detection and segmentation / Open-set and few-shot detection
Handwritten and Printed Text Segmentation: A Signature Case Study
Sina Gholamian, Ali Vahdat
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 four-class formulation with overlap class instead of three-class (ablation) | FCN-light, Weighted CE · WGM-SYN | mean IoU (%): 50 -> 54.02 | Improved |
| use four-class formulation with overlap class instead of three-class (ablation) | FCN-based, Weighted CE · SignaTR6K | mean IoU (%): 81.09 -> 83.15 | Improved |
| add dense CRF post-processing (ablation) | segmentation models · SignaTR6K | mean IoU (%): 84.05 -> 81.17 (4-class FCN, CE) | Hurt |
| add CRF heuristic that only relabels background pixels (ablation) | 4-class FCN-based, CE · SignaTR6K | HT IoU (%): 89.21 -> 91.81 | Improved |
| add Fine Feature Path in parallel with SSP (ablation) | SSP VGG16 · WGM-SYN and SignaTR6K | mean IoU (%): about 7% on WGM-SYN and 8% on SignaTR6K | Improved |
| use VGG16 backbone instead of small FCN-light (ablation) | SSP · WGM-SYN | mean IoU (%): 54.08 -> 47.17 | Hurt |
| use Fusion loss instead of Dice or Weighted Dice on MFM-InceptionV3 (ablation) | MFM-InceptionV3 · WGM-SYN | mean IoU (%): 40.57 -> 58.77 (Dice vs Fusion) | Improved |
| use Fusion loss instead of CE on MFM-InceptionV3 (ablation) | MFM-InceptionV3 · SignaTR6K | mean IoU (%): 88.18 vs 88.15 (CE vs Fusion) | No clear effect |
| best design vs best prior work | MFM with CRFH · WGM-SYN | mean IoU (%): 50 -> 58.93 | Improved |
| best design vs best prior work | MFM with CRFH · SignaTR6K | mean IoU (%): 83.02 -> 89.10 | Improved |
| use multi-label sigmoid formulation instead of four-class softmax (ablation) | segmentation models · text segmentation | IoU | Hurt |
Closest papers by meaning
- Handwritten Text Recognition: A Survey (TPAMI 2025 · similarity 0.86)
- OrigamiNet: Weakly-Supervised, Segmentation-Free, One-Step, Full Page Text Recognition by learning to unfold (CVPR 2020 · similarity 0.85)
- General Detection-based Text Line Recognition (NeurIPS 2024 · similarity 0.84)
- OCRBench v2: An Improved Benchmark for Evaluating Large Multimodal Models on Visual Text Localization and Reasoning (NeurIPS 2025 · similarity 0.82)
- SRFormer: Text Detection Transformer with Incorporated Segmentation and Regression (AAAI 2024 · similarity 0.82)
- ODM: A Text-Image Further Alignment Pre-training Approach for Scene Text Detection and Spotting (CVPR 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.
