Shift-Net: Image Inpainting via Deep Feature Rearrangement

TL;DR

Shift-Net introduces a novel deep learning architecture with a shift-connection layer that significantly improves image inpainting by producing sharper and more detailed results compared to previous methods.

Contribution

The paper proposes Shift-Net, a new U-Net based architecture with a shift-connection layer and guidance loss for improved image inpainting of arbitrary shapes.

Findings

Produces sharper, more detailed inpainted images

Demonstrates effectiveness on Paris StreetView and Places datasets

Outperforms existing inpainting methods

Abstract

Deep convolutional networks (CNNs) have exhibited their potential in image inpainting for producing plausible results. However, in most existing methods, e.g., context encoder, the missing parts are predicted by propagating the surrounding convolutional features through a fully connected layer, which intends to produce semantically plausible but blurry result. In this paper, we introduce a special shift-connection layer to the U-Net architecture, namely Shift-Net, for filling in missing regions of any shape with sharp structures and fine-detailed textures. To this end, the encoder feature of the known region is shifted to serve as an estimation of the missing parts. A guidance loss is introduced on decoder feature to minimize the distance between the decoder feature after fully connected layer and the ground-truth encoder feature of the missing parts. With such constraint, the decoder feature in missing region can be used to guide the shift of encoder feature in known region. An end-to-end learning algorithm is further developed to train the Shift-Net. Experiments on the Paris StreetView and Places datasets demonstrate the efficiency and effectiveness of our Shift-Net in producing sharper, fine-detailed, and visually plausible results. The codes and pre-trained models are available at https://github.com/Zhaoyi-Yan/Shift-Net.