Hadoop OCR

MapReduce job for text recognition from YouTube videos.

Installation

Tested on Ubuntu 14.04, but highly recommend to use Ubuntu 14.10.

• If you are using Ubutnu 14.04 or lower, add Ubuntu 14.10 repository to sources.list. We need it to install latest opencv 2.4.9 libraries.

echo 'deb http://mirrors.kernel.org/ubuntu utopic main universe' >> /etc/apt/sources.list
sudo apt-get update

• Install tesseract

  sudo apt-get install tesseract-ocr

• Install libopencv-dev 2.4.9

  sudo apt-get install libopencv-dev

• Install Hadoop 2.6.0, CDH5 recommended

• Checkout reposotory and build jar with mvn clean install

• Create initial HDFS structure

  sudo -u hdfs hadoop fs -mkdir /user/ubuntu
  sudo -u hdfs hadoop fs -chown ubutnu /user/ubutnu
  

• Create input and set input file

  scripts/set_url.sh <youtube_url>
  

• Run MapReduce job

  scripts/clean.sh && scripts/run.sh <absolute_path_to_jar> && scripts/result.sh