not-shoyo/kannada-digit-ocr

A program that can perform optical character recognition of Kannada digits using the KNN algorithm.

Optical Character Recognition of Kannada digits

Description

This project involves a program that identifies a Kannada digit that is given as input by making a custom dataset to run a lazy trained KNN Machine Learning Algorithm. Another python program is run to convert the input image dataset into MNIST format which is then used for the machine learning process. A paper written for this project can be found here.

Setup

• Download or Clone the project and run ocr.py

• If any error is encountered, check if the packages numpy and imagio have been imported properly in ocr.py and new_data/convert.py .

Usage

• Run ocr.py to run the project.

• The program parses the images in our_input and predicts labels for them. These labels are then compared with the actual labels (known from the file name) of the images to determine the accuracy of the program.

• This program only works on input images that are black and white having a dimension of 28 x 28 pixels.