文本检测算法性能对比

文本检测算法性能对比

  • 1. 任意四边形文本数据集
    • 1.1 ICDAR15[^1]
  • 2. 曲形文本数据集
    • 2.1 CTW1500[^10]
    • 2.2 Total-Text[^11]

本文会一直更新经典的、最新的或性能最好的文本检测算法

1. 任意四边形文本数据集

1.1 ICDAR151

该数据集包含1000张图片,其中训练集500张,测试集500张,这些图片从谷歌街景中搜集。目标是多个方向,标注为word级别的,四个点的坐标。

算法 发表时间 算法类型 P R F
CTPN2 ECCV-2016 Regression 0.74 0.52 0.61
IncepText3 IJCAI-2018 Segmentation 0.938 0.873 0.905
PSENet4 CVPR-2019 Segmentation 0.8692 0.845 0.8569
CRAFT5 CVPR-2019 Segmentation 0.898 0.843 0.869
CRAFTS6 ECCV-2020 Segmentation 0.853 0.890 0.871
EAST7 CVPR-2017 Hybrid 0.833 0.783 0.807
DB8 AAAI-2020 Hybrid 0.918 0.832 0.873
ContourNet9 CVPR2020 Hybrid 0.94 0.901 0.87
DRRG10 CVPR2020 GCN 0.8853 0.8469 0.8656
TextFuseNet11 IJCAI-PRICAI-20 Hybrid 0.940 0.906 0.922
SDM12 ECCV-2020 Segmentation 0.9196 0.8922 0.9057

2. 曲形文本数据集

2.1 CTW150013

该数据集是曲形文本检测集,包含1000张训练集和500张测试集,文本使用14个边界点标注,行标注级别。

算法 发表时间 算法类型 P R F
PSENet4 CVPR-2019 Segmentation 0.848 0.797 0.822
CRAFT5 CVPR-2019 Segmentation 0.86 0.811 0.835
DB8 AAAI-2020 Hybrid 0.869 0.802 0.834
ContourNet9 CVPR2020 Hybrid 0.857 0.84 0.848
DRRG10 CVPR2020 GCN 0.8593 0.8302 0.8445
TextFuseNet11 IJCAI-PRICAI-20 Hybrid 0.897 0.851 0.874
SDM12 ECCV-2020 Segmentation 0.8840 0.8442 0.8636

2.2 Total-Text14

与CTW1500不同的是,标注是word级别的,该数据集包含水平方向、多方向和曲形文本,共1225张训练集和300张测试集图片。

算法 发表时间 算法类型 P R F
PSENet4 CVPR-2019 Segmentation 0.84 0.779 0.809
CRAFT5 CVPR-2019 Segmentation 0.876 0.799 0.836
CRAFTS6 ECCV-2020 Segmentation 0.854 0.895 0.874
DB8 AAAI-2020 Hybrid 0.871 0.825 0.847
ContourNet9 CVPR2020 Hybrid 0.869 0.839 0.854
DRRG10 CVPR2020 GCN 0.8654 0.8493 0.8573
TextFuseNet11 IJCAI-PRICAI-20 Hybrid 0.892 0.858 0.875
SDM12 ECCV-2020 Segmentation 0.9085 0.8603 0.8837

  1. ICDAR2015[70]:D. Karatzas, L. Gomez-Bigorda, A. Nicolaou, S. K. Ghosh, A. D.Bagdanov, M. Iwamura, J. Matas, L. Neumann, V. R. Chandrasekhar, S. Lu, F. Shafait, S. Uchida, and E. Valveny. ICDAR 2015 competition on robust reading. In ICDAR, pages 1156–1160, 2015. Paper ↩︎

  2. Tian Z, Huang W, He T, et al. Detecting text in natural image with connectionist text proposal network. European conference on computer vision(ECCV), 2016: 56-72. Paper Code ↩︎

  3. Qiangpeng Yang, Mengli Cheng et al. IncepText: A New Inception-Text Module with Deformable PSROI Pooling for Multi-Oriented Scene Text Detection. In IJCAI 2018. Paper ↩︎

  4. Wenhai W, Enze X, et al. Shape Robust Text Detection with Progressive Scale Expansion Network. In CVPR 2019. Paper Code ↩︎ ↩︎ ↩︎

  5. Youngmin Baek, Bado Lee, et al. Character Region Awareness for Text Detection. In CVPR 2019. Paper ↩︎ ↩︎ ↩︎

  6. Baek Y , Shin S , Baek J , et al. Character Region Attention For Text Spotting[J]. 2020. ↩︎ ↩︎

  7. Zhou X, Yao C, Wen H, et al. EAST: an efficient and accurate scene text detector. CVPR, 2017: 2642-2651. Paper Code ↩︎

  8. Minghui Liao, et al, Real-time Scene Text Detection with Differentiable Binarization. In AAAI, 2020. PaperCode ↩︎ ↩︎ ↩︎

  9. Wang Y , Xie H , Zha Z , et al. ContourNet: Taking a Further Step toward Accurate Arbitrary-shaped Scene Text Detection[C]// 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2020. ↩︎ ↩︎ ↩︎

  10. Zhang S X , Zhu X , Hou J B , et al. Deep Relational Reasoning Graph Network for Arbitrary Shape Text Detection[J]. 2020. ↩︎ ↩︎ ↩︎

  11. Ye J , Chen Z , Liu J , et al. TextFuseNet: Scene Text Detection with Richer Fused Features[C]// Twenty-Ninth International Joint Conference on Artificial Intelligence and Seventeenth Pacific Rim International Conference on Artificial Intelligence {IJCAI-PRICAI-20. 2020. ↩︎ ↩︎ ↩︎

  12. Xiao S , Peng L , Yan R , et al. Sequential Deformation for Accurate Scene Text Detection[M]// Computer Vision – ECCV 2020. 2020. ↩︎ ↩︎ ↩︎

  13. Yuliang L, Lianwen J, Shuaitao Z, et al. Curved Scene Text Detection via Transverse and Longitudinal Sequence Connection. Pattern Recognition, 2019.Paper ↩︎

  14. Chee C K, Chan C S. Total-text: A comprehensive dataset for scene text detection and recognition.Document Analysis and Recognition (ICDAR), 2017 14th IAPR International Conference on. IEEE, 2017, 1: 935-942.Paper ↩︎

你可能感兴趣的:(文本检测,目标检测,OCR,深度学习,神经网络,算法,人工智能,python)