Research map / Detection and segmentation / CNN features
Efficient and Accurate Arbitrary-Shaped Text Detection With Pixel Aggregation Network
Wenhai Wang, Enze Xie, Xiaoge Song, Yuhang Zang, Wenjia Wang, Tong Lu, Gang Yu, Chunhua Shen
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 one FPEM (n_c from 0 to 1) (ablation) | PAN · ICDAR 2015 | F-measure (%): 78.4 -> 79.9 | Improved |
| add one FPEM (n_c from 0 to 1) (ablation) | PAN · CTW1500 | F-measure (%): 78.8 -> 80.4 | Improved |
| increase cascaded FPEMs from 2 to 4 (ablation) | PAN · CTW1500 | F-measure (%): 81.0 -> 81.5 | No clear effect |
| increase cascaded FPEMs from 0 to 4 (ablation) | PAN · CTW1500 | FPS: 49.7 -> 32.4 | Hurt |
| add FFM instead of using only the last pyramid (ablation) | PAN · ICDAR 2015 | F-measure (%): 79.7 -> 80.3 | Improved |
| add FFM instead of using only the last pyramid (ablation) | PAN · CTW1500 | F-measure (%): 80.2 -> 81.0 | Improved |
| use FFM instead of direct concatenation (ablation) | PAN · CTW1500 | FPS: 35.9 -> 39.8 | Improved |
| add pixel aggregation (PA) (ablation) | PAN · ICDAR 2015 | F-measure (%): 79.3 -> 80.3 | Improved |
| add pixel aggregation (PA) (ablation) | PAN · CTW1500 | F-measure (%): 79.8 -> 81.0 | Improved |
| replace ResNet18 backbone with ResNet50 (ablation) | PAN · ICDAR 2015 | F-measure (%): 80.3 -> 81.4 | Mixed |
| replace ResNet18 backbone with VGG16 (ablation) | PAN · ICDAR 2015 | F-measure (%): 80.3 -> 81.9 | Mixed |
| replace ResNet50 + PSPNet with ResNet18 + 2 FPEMs + FFM (ablation) | PAN · ICDAR 2015 | FPS: 4.6 -> 26.1 | Improved |
Closest papers by meaning
- PAN++: Towards Efficient and Accurate End-to-End Spotting of Arbitrarily-Shaped Text (TPAMI 2021 · similarity 0.90)
- Towards Unconstrained End-to-End Text Spotting (ICCV 2019 · similarity 0.90)
- Arbitrary Shape Text Detection via Segmentation with Probability Maps (TPAMI 2022 · similarity 0.89)
- Towards Unified Multi-granularity Text Detection with Interactive Attention (ICML 2024 · similarity 0.87)
- Kernel Adaptive Convolution for Scene Text Detection via Distance Map Prediction (CVPR 2024 · similarity 0.87)
- LRANet: Towards Accurate and Efficient Scene Text Detection with Low-Rank Approximation Network (AAAI 2024 · similarity 0.86)
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.
