Hindi optical character recognition

Installation and setup

• If you are cloning locally then install dependencies by running:

pip3 install -r requirements.txt

• If you are using colab then install additional dependencies by running:

sh setup.sh

Dataset

• The dataset can be found here https://archive.ics.uci.edu/ml/datasets/Devanagari+Handwritten+Character+Dataset
• After extracting files run setup.sh with variables TRAIN_PATH and TEST_PATH having their respective paths. This is to remove classes containg digits which we wont train on.

A batch of training images

File Description

Edit files and execute files in the following order and change paths where ever necessary

• utils.py - contains all the necessary plot, load and save model helper functions.

• config.py - contains all the necessary hyperparameters and paths to dataset and also the character to numeric mapping.

• pre_csv.py - creates a csv file for train and test, which contains the path to images along with its label.

• dataloader.py - implements augmentation and preproceesing on custom dataset and clubs images into batches using dataloader.

• model.py (used for experimentaiton) - contains pretrained and custom models which were used for experimentaiton. We finally decided to use ResNet34 pretrained on Imagenet dataset, which gave a 99.75 validation accuracy.

• train.py - train model

• remove.sh - removes numberic character images from dataset extracted

• setup.sh - run it if you are using colab to run the files