Skip to content

A deep learning project using TensorFlow/Keras to clean and denoise dirty document images. Built as a university course project based on the Kaggle "Denoising Dirty Documents" competition dataset. Features a U-Net style autoencoder architecture with mixed precision training for efficient image restoration.

Notifications You must be signed in to change notification settings

Ranjit2111/Denoising-Dirty-Image-Documents

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Denoising Dirty Document Images

This repository contains code and datasets used for the task of denoising dirty document images as well as a PPT giving an overview of the process. I did this as a project for my university course.

Project Overview

The dataset used for this project is based on a Kaggle competition. Due to file size restrictions on GitHub, I will upload the three zip files contained within the main dataset zip file individually. The competition link can be found here.

The primary objective of this project is to build a robust convolutional autoencoder model that can effectively denoise images, facilitating better readability and interpretation of the documents.

Repository Structure

/denoising-dirty-documents
│
├── /train.zip               # Dirty images used for training
├── /train_cleaned.zip       # Cleaned images used for training
├── /test.zip                # Test images to make predictions
├── DenoisingCAE.ipynb       # Jupyter notebook used for the process.

Acknowledgements

Special thanks to Kaggle for providing the dataset, and to the open-source community for the resources and libraries that made this project possible.

About

A deep learning project using TensorFlow/Keras to clean and denoise dirty document images. Built as a university course project based on the Kaggle "Denoising Dirty Documents" competition dataset. Features a U-Net style autoencoder architecture with mixed precision training for efficient image restoration.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published