Code for Handwriting Recognition Course 2020 RUG

Authors:

• Hari Vidharth - s4031180
• Krishnakumar Santhakumar - s4035992
• Dhawal Salvi - s4107624
• Ashwin Vaidya - s3911888

The full code for the pipeline is in the main.py code. The test documents are given as the input one at a time, which then generates the lines folder which contains the segmented lines. The segmented_characters generated folder contains the characters which are segmented from the lines. The next part of the system uses the segmented images from the segmented_characters generated folder and using the saved model in the models folder to perform character recognition on the segmented images and the output is in the form of document_name.txt containing the transcribed text. The final part of the system also uses the segmented images from the segmented_characters generated folder and using the saved model in the Models folder to perform the style classification on the segmented images and the output is in the form of document_name_StyleOutput.txt containing the period/era text.

Make sure that the pip has been upgraded so that latest version on tensorflow is installed. Also ensure that your are using Python 3. pip3 install --upgrade pip

Then install the requirements pip3 install -r requirements.txt

To run the code follow the command and enter the folder which contains the images of the documents. Make sure to install the dependencies mentioned in requirements.txt

python3 main.py --image path/to/image_folder

python3 main.py --image ./test_images

line_segmentator module contains two scripts. They are used to segment the given document into lines. utils.py contains code which aids is finding rotation on the document, calculating projection profile and A*.

character_segmentor contains two scripts. The ConnectedComponents.py does a rough segmentation on an image of a line into characters and words. The template_matching.py script does template matching on the rough segmentation to get a more fine grained segmentation.

character_recognition module contains script to recognize the segmented characters.

style_classification module contains styleClassificationTest script to perform the final period classification using the segmented characters.