In this paper, we propose a multi-stage and high-resolution model for imagesynthesis that uses fine-grained attributes and masks as input . With mask as prior, the model in this paper isconstrained so that the generated images conform to visual senses . This paper also proposes a scheme to improve thediscriminator of the generative adversarial network by simultaneouslydiscriminating the total image and sub-regions of the image . In addition, we also propose a method for optimizing the labeled attribute in datasets, which reduces the manual labeling noise . Extensive quantitative results show that ourimage synthesis model generates more realistic images .

Author(s) : Pengyang Li, Donghui Wang

Links : PDF - Abstract

Code :
Coursera

Keywords : image - paper - model - images - propose -

Leave a Reply

Your email address will not be published. Required fields are marked *