Documentation:

Content:

1. Installation
2. Download dataset
3. Train and Test HMM
4. Start Server
5. Output

1 Installation

1.1 Copy contents

	Copy contents onto local disk

1.2 Install Python 2.7 

	https://www.python.org/download/releases/2.7/

1.3 Install OpenCV 2.7, seqlearn

	OpenCV 2.7

	Windows: http://docs.opencv.org/2.4/doc/tutorials/introduction/windows_install/windows_install.html

	Linux: http://docs.opencv.org/2.4/doc/tutorials/introduction/linux_install/linux_install.html

1.4 Install PIP and its libraries

	numpy, scipy, sklearn, nltk, gensim, wordcloud, collections, PIL

	seqlearn

		Clone repository: https://github.com/larsmans/seqlearn

		Get NumPy >=1.6, SciPy >=0.11, Cython >=0.20.2 and a recent version of scikit-learn. Then issue:

		python setup.py install
		to install seqlearn.

		If you want to use seqlearn from its source directory without installing, you have to compile first:

		python setup.py build_ext --inplace

1.5 Install NodeJS

	https://nodejs.org/en/download/package-manager/

1.6 Install NodeJS packages

	Go to root directory of project and type "npm install"

2 Datasets

2.1 Register http://www.fki.inf.unibe.ch/DBs/iamDB/iLogin/index.php

2.2 Download http://www.fki.inf.unibe.ch/DBs/iamDB/data/forms (FormsA-E, FormsE-H, FormsI-Z

2.3 Download http://www.fki.inf.unibe.ch/DBs/iamDB/data/words

2.4 Run "python processdata.py" in "python" folder

3 Train and Test HMM

3.1 Navigate to "python" folder

3.2 Type "python train.py"

3.3 Type "python test.py"

4 Start Server

"node server.js"

5 Output

Open browser and navigate to localhost:3000

Upload a Handwritten document with multiple lines and click on "Extract"

Navigate to different tabs (Beautify, Recognize, Summarize) and see the desired result

5.1 Training output:
	The output is a trained classifier which will be used in the web interface

5.2 Beautification output:
	Can be found in the "Beautify"	

5.3 Recognition output:
	Can be found in the "Recognize". Depends on training data.

5.4 Summarize output:
	Can be found in the "Summarize". Here, the category of text must be selected.