Skip to content

This project focuses on Corrupted Image Modeling (CiM), where the goal is to train a model capable of reconstructing or enhancing images with missing or corrupted regions. The approach involves learning a robust representation of images to infer and restore missing details.

Notifications You must be signed in to change notification settings

Aditya-1874/Corrupted-Image-Modelling

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Corrupted Image Modeling

Overview

This project focuses on Corrupted Image Modeling (CiM), where the goal is to train a model capable of reconstructing or enhancing images with missing or corrupted regions. The approach involves learning a robust representation of images to infer and restore missing details.

Methodology

  • Utilizes a BEiT-based masked image modeling approach, where images are divided into patches, and specific patches are masked out.
  • The model is trained to predict the masked patches using self-supervised learning, enabling it to learn meaningful representations of image structures.
  • A DALL-E decoder is used to reconstruct the missing parts based on learned embeddings.

Progress

  • Successfully implemented the generator model, which processes corrupted images and produces meaningful reconstructions.
  • Currently working on integrating a more refined enhancement module to improve restoration quality.

Future Work

  • Optimize the reconstruction pipeline by experimenting with different architectures.
  • Evaluate performance on various datasets to measure the model’s effectiveness.
  • Extend the model for real-world applications such as medical imaging, satellite image restoration, and digital forensics.

Reference

This project is inspired by BEiT: BERT Pre-Training of Image Transformers. You can read more about it here:
🔗 BEiT Paper (arXiv:2202.03382)

Usage

To be updated once the full pipeline is complete.

About

This project focuses on Corrupted Image Modeling (CiM), where the goal is to train a model capable of reconstructing or enhancing images with missing or corrupted regions. The approach involves learning a robust representation of images to infer and restore missing details.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published