Research map / Detection and segmentation / CNN features
Towards Unconstrained End-to-End Text Spotting
Siyang Qin, Alessandro Bissacco, Michalis Raptis, Yasuhisa Fujii, Ying Xiao
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 RoI masking (no partially labeled data) (ablation) | E2E-baseline, Inception-ResNet · ICDAR15 test | AP end-to-end: 51.7 -> 53.1 | Improved |
| add RoI masking (no partially labeled data) (ablation) | E2E-baseline, ResNet-50 · ICDAR15 test | AP end-to-end: 52.8 -> 53.9 | Improved |
| add RoI masking on top of partially labeled data (ablation) | E2E, ResNet-50 · ICDAR15 test | AP end-to-end: 55.7 -> 59.0 | Improved |
| add RoI masking on top of partially labeled data (ablation) | E2E, Inception-ResNet · ICDAR15 test | AP end-to-end: 58.7 -> 59.5 | Improved |
| add partially labeled data (ablation) | E2E-baseline, ResNet-50 · ICDAR15 test | AP end-to-end: 52.8 -> 55.7 | Improved |
| add partially labeled data (ablation) | E2E-baseline, Inception-ResNet · ICDAR15 test | AP end-to-end: 51.7 -> 58.7 | Improved |
| add partially labeled data (ablation) | E2E-baseline, Inception-ResNet · ICDAR15 test | AP detection: 88.2 -> 89.9 | Improved |
| single-step instead of two-step training (ablation) | E2E-full, ResNet-50 · ICDAR15 test | AP end-to-end: 55.3 -> 59.0 | Improved |
| single-step instead of two-step training (ablation) | E2E-full, Inception-ResNet · ICDAR15 test | AP end-to-end: 57.4 -> 59.5 | Improved |
| add partially labeled data (public-only vs full training data) (ablation) | Ours Inception-ResNet · Total-Text | End-to-end F-score: 63.9 -> 70.7 | Improved |
| use Inception-ResNet instead of ResNet-50 backbone (ablation) | Ours · Total-Text | End-to-end F-score: 67.8 -> 70.7 | Improved |
| use Inception-ResNet instead of ResNet-50 backbone, no partially labeled data (ablation) | E2E-baseline · ICDAR15 test | AP end-to-end: 52.8 -> 51.7 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FOTS — FOTS: Fast Oriented Text Spotting With a Unified Network (CVPR 2018) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- All You Need Is Boundary: Toward Arbitrary-Shaped Text Spotting (AAAI 2020 · similarity 0.93)
- Text Perceptron: Towards End-to-End Arbitrary-Shaped Text Spotting (AAAI 2020 · similarity 0.92)
- TextDragon: An End-to-End Framework for Arbitrary Shaped Text Spotting (ICCV 2019 · similarity 0.92)
- Mask TextSpotter: An End-to-End Trainable Neural Network for Spotting Text with Arbitrary Shapes (ECCV 2018 · similarity 0.91)
- Towards End-to-End Text Spotting in Natural Scenes (TPAMI 2021 · similarity 0.90)
- LRANet: Towards Accurate and Efficient Scene Text Detection with Low-Rank Approximation Network (AAAI 2024 · similarity 0.87)
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.
