ML Project - Handwriting to Digital Text Conversion (H2DTC)

Overview

This repository contains the code and documentation for the Handwriting to Digital Text Conversion (H2DTC) project, developed as part of CS550, Project Phase 1 at IIT Bhilai. The projct code can found in the ML Final.ipynb file and the HMM model can be found under hmm_model.ipynb file. Report can be found under ML_Project Report.pdf.

Team Members

• JV Aditya (12140840)
• C Nikhil (12140530)
• B Sri Bhargav Ram (12140460)

Project Description

1. Data Pre-processing

We utilized the MNIST and EMNIST datasets, comprising over 400,000 handwritten words for character recognition. The data was divided into training (331,059), testing (41,382), and validation sets (41,382). The preprocessing involved flattening images into a 1D array and using Principal Component Analysis (PCA) for dimensionality reduction (64 components).

A sample images from the training and validation dataset:

A sample image from the testing dataset:

2. Models Used

Our project successfully identified handwriting in images, achieving a commendable accuracy of approximately 93%. We conducted extensive experiments, exploring various models such as Convolutional Neural Networks (CNN)-7 models used,Resnet -3 models used Support Vector Machines (SVM), K-Means clustering, and Hidden Markov Models (HMMs) for character recognition.We have successfully implemented an ensemble approach using the three models that yielded the highest accuracy. This ensemble leverages the predictive power of the top-performing models to enhance the overall performance, resulting in a more robust and accurate system

3. Individual Contributions

• J V Aditya: Pre-processing and Image Enhancement,SVM Model,Grapheme Segmentation,Ensembling.
• Chiruvolu Nikhil: Feature Extraction,7 different CNN models,Hidden Markov Models.
• Bollapragada Sri Bhargav: 3 different ResNet Models,. C-RNN Networks,K-Means Clustering