Image Completion on CIFAR-10

TL;DR

This paper explores image completion on CIFAR-10 using various neural network architectures, identifying the most effective model that produces realistic in-painted images with low error.

Contribution

It compares three neural network architectures for image completion on CIFAR-10 and identifies the deep fully convolutional network as the most effective.

Findings

Deep fully convolutional network achieved MSE of 0.015.

In-painted images appeared realistic to humans.

Model outperformed other architectures in quality.

Abstract

This project performed image completion on CIFAR-10, a dataset of 60,000 32x32 RGB images, using three different neural network architectures: fully convolutional networks, convolutional networks with fully connected layers, and encoder-decoder convolutional networks. The highest performing model was a deep fully convolutional network, which was able to achieve a mean squared error of .015 when comparing the original image pixel values with the predicted pixel values. As well, this network was able to output in-painted images which appeared real to the human eye.