O-T-O-Z/Bachelor-s-project-s3951731

Repository containing files related to my bachelor's thesis.

Introduction

This codebase represents the scripts that were used in the bachelor's thesis called "Siamese Convolutional Neural Networks In Handwriting Verification Using Word Texts". Each file is explained below.

Model architecture

The model architecture can be observed below.

Example training

An example of a training session can be seen below.

Code

data.zip

Contains all the pre-processed data that was used in this project:

• CVL_100: Samples originating from the CVL database (downloaded from https://cvl.tuwien.ac.at/research/cvl-databases/an-off-line-database-for-writer-retrieval-writer-identification-and-word-spotting/)
• IAM_100: Samples originating from the IAM database (downloaded from https://fki.tic.heia-fr.ch/databases/iam-handwriting-database)
• Mixed_padded: Mixed database containing all samples from CVL_100 and IAM_100.

grayscaling.py

Script that converts images in a folder to grayscale format.

padding.py

Script that resizes and adds white padding pixels to images to convert them to square images of 256 by 256.

test.py

Tests a certain trained model on a certain testing dataset of the same form.

train.py

Trains a certain model on a training dataset containing a training and validation set of the same form.