DeepGIN: Deep Generative Inpainting Network for Extreme Image Inpainting

TL;DR

DeepGIN is a versatile deep generative network designed for extreme image inpainting, effectively handling various missing patterns with innovative modules like SPD ResNet, MSSA, and BP, outperforming existing methods.

Contribution

The paper introduces DeepGIN, a novel inpainting network with specialized modules for diverse missing patterns, advancing the state-of-the-art in image completion.

Findings

Outperforms existing inpainting methods on FFHQ and Oxford Buildings datasets.

Capable of completing masked images in real-world scenarios.

Uses novel SPD ResNet, MSSA, and BP modules to enhance reconstruction quality.

Abstract

The degree of difficulty in image inpainting depends on the types and sizes of the missing parts. Existing image inpainting approaches usually encounter difficulties in completing the missing parts in the wild with pleasing visual and contextual results as they are trained for either dealing with one specific type of missing patterns (mask) or unilaterally assuming the shapes and/or sizes of the masked areas. We propose a deep generative inpainting network, named DeepGIN, to handle various types of masked images. We design a Spatial Pyramid Dilation (SPD) ResNet block to enable the use of distant features for reconstruction. We also employ Multi-Scale Self-Attention (MSSA) mechanism and Back Projection (BP) technique to enhance our inpainting results. Our DeepGIN outperforms the state-of-the-art approaches generally, including two publicly available datasets (FFHQ and Oxford Buildings), both quantitatively and qualitatively. We also demonstrate that our model is capable of completing masked images in the wild.