Research map / Detection and segmentation / CNN features
Detecting Tampered Scene Text in the Wild
Yuxin Wang, Hongtao Xie, Mengting Xing, Jing Wang, Shenggao Zhu, Yongdong Zhang
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 |
|---|---|---|---|
| separate the segmentation branch (S3R) | EAST · Tampered-IC13 test | mF: 50.39 -> 52.70 | Improved |
| separate the segmentation branch (S3R) | PSENet · Tampered-IC13 test | mF: 37.46 -> 64.76 | Improved |
| separate the segmentation branch (S3R) | ContourNet · Tampered-IC13 test | mF: 75.98 -> 76.66 | Improved |
| share the regression (GP) process (ablation) | EAST · Tampered-IC13 test | mF: 31.18 -> 52.70 | Improved |
| separate GP heads without representation suppression (ablation) | ContourNet · Tampered-IC13 test | mF: 75.98 -> 74.18 | Hurt |
| add representation suppression on last-layer segmentation maps with shared GP head (ablation) | ContourNet · Tampered-IC13 test | mF: 75.98 -> 76.16 | Improved |
| add representation suppression to separated GP heads (ablation) | ContourNet · Tampered-IC13 test | mF: 74.18 -> 76.66 | Improved |
| kernel size 7x7 vs 5x5 in parallel-branch extractor (ablation) | EAST (SynthText pretrained) · Tampered-IC13 test | mF: 60.03 -> 62.96 | Improved |
| kernel size 9x9 vs 7x7 in parallel-branch extractor (ablation) | EAST (SynthText pretrained) · Tampered-IC13 test | mF: 62.96 -> 62.33 | Hurt |
| add parallel-branch (frequency) feature extractor | EAST (SynthText pretrained) · Tampered-IC13 test | mF: 60.39 -> 62.96 | Improved |
| add parallel-branch (frequency) feature extractor | ATRR · Tampered-IC13 test | mF: 75.68 -> 76.35 | Improved |
| add parallel-branch extractor when training on half the images | EAST (SynthText pretrained) · Tampered-IC13 test (half training images) | mF: 57.87 -> 60.79 | Improved |
Closest papers by meaning
- Revisiting Tampered Scene Text Detection in the Era of Generative AI (AAAI 2025 · similarity 0.89)
- TextShield-R1: Reinforced Reasoning for Tampered Text Detection (AAAI 2026 · similarity 0.86)
- OTE: Exploring Accurate Scene Text Recognition Using One Token (CVPR 2024 · similarity 0.85)
- Toward Real-world Text Image Forgery Localization: Structured and Interpretable Data Synthesis (NeurIPS 2025 · similarity 0.85)
- Boosting Semi-Supervised Scene Text Recognition via Viewing and Summarizing (NeurIPS 2024 · similarity 0.84)
- FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models (ICLR 2025 · 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.
