Cursive Handwriting Recognition (Incomplete)

This repository is for training a cursive handwriting model for Image and Video Recognition course final project at TITech 2020 4Q.
This model was trained on Kaggle A-Z dataset and further transfer learned using custom cursive dataset.
Even though this model can recognize individual letters and can detect cursive letters using sliding window algorithm, this model DOES NOT WORK on cursive handwritten words.

Getting started

Edit the image filename in the

handwriting_recognition.py

file and run. This file uses pretrained model. If you want to train using on your own dataset edit files:

train_base_CNN_model.py
train_transfer_model.py

Prerequisites

tensorflow 2.4
pytessaract
opencv2