DeepText: A Python repository from zhongzhuoyao

DeepText: A new approach for text proposal generation and text detection in natural images.

by Zhuoyao Zhong, Lianwen Jin, Shuangping Huang, South China University of Technology (SCUT), Published in ICASSP 2017.

Introduction

This repository is a fork from py-faster-rcnn, and our proposed DeepText system for scene textdetection is based on the elegant framework of Faster R-CNN.

You can refer to py-faster-rcnn README.md and faster-rcnn README.md for more information.

Desclaimer

Please note that this repository is the demo codes (with our trained model) for DeepText system, which doesn't contain iterative regression module and linking segments as well as any training codes.

Citing DeepText

If our codes are useful for your work, please cite our paper:

@inproceedings{icassp2017DeepText,
  title={{DeepText}: DeepText: A new approach for text proposal generation and text detection in natural images},
  author={Zhuoyao Zhong, Lianwen Jin, Shuangping Huang},
  booktitle = {International Conference on Acoustics, Speech and Signal Processing ({ICASSP})}},
  year={2017}
}

Installation

Clone the DeepText repository

# Make sure to clone with --recursive
git clone --recursive https://github.com/zhongzhuoyao/DeepText.git

We'll call the directory that you cloned DeepText into DeepText_ROOT. Build the Cython modules
```
cd $DeepText_ROOT/lib
make
```

Build Caffe and pycaffe

cd $DeepText_ROOT/caffe-fast-rcnn
# Now follow the Caffe installation instructions here:
#   http://caffe.berkeleyvision.org/installation.html
# For your Makefile.config:
#   Uncomment `WITH_PYTHON_LAYER := 1`

cp Makefile.config.example Makefile.config
make -j8 && make pycaffe

Download DeepText text detection model from one drive, and then populate it into directory models/text_detection/. The model's name should be vgg16_DeepText_trained_model.caffemodel.

How to run the demo

You can download ICDAR-2013 benchmark or any text images and then populate them into directory demo_text_images.
Run python tools/demo_DeepText.py under GPU mode or python tools/demo_DeepText.py --cpu to run it under CPU mode.

Expected results

Recall, Precision and F-measure on the ICDAR-2013 benchmark.

Recall (%)	Precision (%)	F-measure (%)
82.17	87.13	84.58