Transformer with Convolution for Irregular Image Inpainting

In recent years, image inpainting is a challenging task in the field of computer vision. Many methods based on deep learning have been proposed to deal with the problem of image inpainting, but most of these methods are based on u-net. And recently, the transformer has demonstrated its powerful capa...

Full description

Saved in:
Bibliographic Details
Published in2022 International Conference on Image Processing, Computer Vision and Machine Learning (ICICML) pp. 35 - 38
Main Authors Xie, Donglin, Wang, Lingfeng
Format Conference Proceeding
LanguageEnglish
Published IEEE 28.10.2022
Subjects
Computational modeling
Computer vision
Convolution
Deep learning
image inpainting
Image processing
irregular
Task analysis
transformer
Transformers
Online AccessGet full text

Cover

More Information
Summary:In recent years, image inpainting is a challenging task in the field of computer vision. Many methods based on deep learning have been proposed to deal with the problem of image inpainting, but most of these methods are based on u-net. And recently, the transformer has demonstrated its powerful capabilities in the field of computer vision. In this paper, we propose a transformer-based image inpainting model to solve the problem of image inpainting with irregular missing regions. In particular, we utilize partial convolution to generate patch embedding, avoiding the influence of missing region pixels. The proposed model combines transformer and convolution to simultaneously focus on global and local features. Finally, experimental results on public datasets show that our model outperforms traditional convolutional inpainting models.
DOI:10.1109/ICICML57342.2022.10009791