Experince Future with PenPixelPioneers

Beyond Pen and Pixel:
A Symphony of Styles, A World of Predictions

How It Works

Get to Know the Project

Our project is a desktop app that's able to analyze and understand your handwriting. Users can add training data to teach the app about personal writing styles. Our image recognition system is able analyze and recognize the writer up to 90%.

Users can also create their own fonts to simulate their handwriting. By filling the template that we created users are able to save their custom fonts and create texts with their fonts. This text is sent to arduino to recreate the users writing style.

Image Recognition

Custom Fonts

Arduino Integration

Our desktop app uses image recognition algorithms to identify writers from test images. You can then use our custom font creator to make your own unique fonts for handwritten text simulations. Finally, you can send your text data to an Arduino to simulate handwriting.

Handwriting Style Imitation and Recognition System

Desktop Application

It has features for adding new user profiles to train it's image recognition mode. Users can also add new font files to save their custom .ttf files. With these setups the users can scan any text sample and get predictions. Also write their own text to be generated and sent to arduino for simulation.

Handwriting style imitation

In this module, users can store the collected dataset of handwritten samples for every desired person. After collecting each person's handwritten style template, crop each character in the template using OpenCV and save these characters in “.png” format in a folder structure. When all the characters of a person are cropped, we convert them to “.png” formatted images to “.bmp” and then to “.svg” format using OpenCV. When SVG files were created, we used FonForge to create “.ttf” files.

Custom created font

Easy to use font template

Writing Style Recognition

In this module, we focused on identifying the writer of a handwritten text using a texture-based approach. The module consists of three main components: pre-processing, feature extraction, and classification. It processes test case images, removes noise, detects handwritten text regions, segment paragraphs into lines, extract texture-based features using LBP, and employ KNN to classify and identify the writer.

Local Binary Pattern Descriptor

K-Nearest Neighbors Classification

Hardware

The coordinates obtained from the computer will be transferred to the Arduino, which will be used to control three-step motors moving along the x and y axes and a servo motor for the Z axis. We also have a black box for better-quality images. With our setup, we can get good scans consistently.