Custom_EAST

EAST paper implementation repository modified for custom detection tasks.

This is an implementation of research paper: https://arxiv.org/abs/1704.03155v2 The code was, actually, written by ArgMan, check out his repository: https://github.com/argman/EAST

This version is modified implementation of EAST to include the calculation of fps on images i.e. the fps on cpu and gpu will also be calculated.

Dependencies

Following dependencies are required to run this.

1. Install Anaconda package from https://docs.anaconda.com/anaconda/install/ for python3.
2. Create a conda environment conda create -n tensorflow3 python=3.5.
3. Install tensorflow v1.3 pip3 install tensorflow.
4. You also need resnet checkpoint to load from. Download it from here: http://download.tensorflow.org/models/resnet_v1_50_2016_08_28.tar.gz

Working

1. To use this, clone or zip download the repository.
2. Give input dataset path and output directory in the eval.py file.
3. The results, coordinates of boxes and boxes on the detected text in images, will be in your output directory.

Running on CPU

1. To run this on CPU, just type python3 eval.py |tee out. Use |tee out to store output in out file for viewing.

Running on GPU

1. To run this on GPU, you need to install and then add CUDA's path in ~/.bashrc file in. Add these lines to .bashrc file:

    export PATH=/usr/local/cuda/bin:${PATH}
    export MANPATH=/usr/local/cuda/man:${MANPATH}
    if [[ "${LD_LIBRARY_PATH}" != "" ]]
    then
      export LD_LIBRARY_PATH=/usr/local/cuda/lib64:${LD_LIBRARY_PATH}
    else
      export LD_LIBRARY_PATH=/usr/local/cuda/lib64
    fi
   

2. Specify the GPU in eval.py file.

3. Type:

source ~/.bashrc
export CUDA_VISIBLE_DEVICES=GPU_Number
python3 eval.py |tee out