ITN: Geometry-Aware Scene Text Detection

Fangfang Wang*, Liming Zhao*, Xi Li, Xinchao Wang, and Dacheng Tao. “Geometry-Aware Scene Text Detection with Instance Transformation Network.” In CVPR, 2018. (*authors contributed equally)

@InProceedings{Wang_2018_CVPR,
author = {Wang, Fangfang and Zhao, Liming and Li, Xi and Wang, Xinchao and Tao, Dacheng},
title = {Geometry-Aware Scene Text Detection with Instance Transformation Network},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
pages = {1381-1389},
year = {2018}
}

Contact:

Fangfang Wang (fangfangliana@gmail.com);

Liming Zhao (http://www.zhaoliming.net/, zlmzju@gmail.com);

Xi Li (http://person.zju.edu.cn/xilics, xilizju@zju.edu.cn)

Instructions

• Use custom Caffe for using transformation layer.

• Run train.sh in the experiments folder to train the model.

Descriptions

• Use Caffe for implementation, please refer to the Caffe project website for installation.

• The protocal files are located in experiments/xxx/proto folder.

• Trained models can be downloaded from google drive or onedrive, and then put them in experiments/xxx/weight folder.