MVA - DELIRES Project - Uncertainty Quantification in Deep Learning for Inverse Problems

MVA - Project for the course Deep Learning for Image Restoration and Synthesis

Author: Clément Bonet

Based on the paper "Uncertainty Quantification in Deep Learning for Inverse Problems" by Alex Kendall and Yarin Gal

Project done with Python.

I used two types of network:

Autoencoder for the Missing Pixel Problem, and the Deblurring Problem: you can see the results in ./Missing Pixels - Deblurring/Results.ipynb
SRCNN for the SISR x2 Problem, inpired by https://github.com/MarkPrecursor/SRCNN-keras. You can see the results in ./SISR/Results.ipynb

Requirements

The project requires some Python libraries: