Research map / Detection and segmentation / CNN features
Sequential Deformation for Accurate Scene Text Detection
Shanyu Xiao, Liangrui Peng, Ruijie Yan, Keyu An, Gang Yao, Jaesik Min
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 the Sequential Deformation Module (ablation) | Mask R-CNN baseline (ResNet-18) · ICDAR 2017 MLT | F-measure (%): 74.84 -> 75.62 | Improved |
| add the Sequential Deformation Module (ablation) | Mask R-CNN baseline (ResNet-50) · ICDAR 2017 MLT | F-measure (%): 77.07 -> 77.62 | Improved |
| add auxiliary character counting to SDM (ablation) | SDM (ResNet-18) · ICDAR 2017 MLT | F-measure (%): 75.62 -> 76.00 | Improved |
| add auxiliary character counting to SDM (ablation) | SDM (ResNet-50) · ICDAR 2017 MLT | F-measure (%): 77.62 -> 78.08 | Improved |
| use character recognition instead of counting as auxiliary task (ablation) | SDM (ResNet-50) · ICDAR 2017 MLT | F-measure (%): 77.62 -> 77.64 | No clear effect |
| increase sequential sampling iteration number T up to 5 (ablation) | SDM (ResNet-18) · ICDAR 2017 MLT | F-measure (%): F-measure firstly increases and then saturates for T >= 5 | Improved |
| auxiliary counting with T = 0 (single feature vector) (ablation) | SDM (ResNet-18) · ICDAR 2017 MLT | F-measure (%): improvement of 0.4% | Improved |
| add SDM and counting to the baseline | Mask R-CNN (ResNet-50) · Total-Text | F-measure (%): 86.16 -> 86.91 | Improved |
| add SDM to the baseline | Mask R-CNN (ResNet-50) · CTW1500 | F-measure (%): 83.06 -> 84.01 | Improved |
| use multi-scale testing | Ours (ResNet-50) · ICDAR 2017 MLT | F-measure (%): 78.08 -> 80.61 | Improved |
| use multi-scale testing | Ours (ResNet-50) · ICDAR 2015 | F-measure (%): 88.57 -> 90.57 | Improved |
| use multi-scale testing | Ours (ResNet-50) · Total-Text | F-measure (%): 86.91 -> 88.37 | Improved |
Closest papers by meaning
- Character Region Awareness for Text Detection (CVPR 2019 · similarity 0.89)
- Arbitrary Shape Scene Text Detection With Adaptive Text Region Representation (CVPR 2019 · similarity 0.89)
- EAST: An Efficient and Accurate Scene Text Detector (CVPR 2017 · similarity 0.88)
- Multi-Oriented Scene Text Detection via Corner Localization and Region Segmentation (CVPR 2018 · similarity 0.88)
- Look More Than Once: An Accurate Detector for Text of Arbitrary Shapes (CVPR 2019 · similarity 0.88)
- ContourNet: Taking a Further Step Toward Accurate Arbitrary-Shaped Scene Text Detection (CVPR 2020 · similarity 0.88)
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.
